Mubarak Shah

Muhammad Awais, Muzammal Naseer, Salman Khan et al.

Vision systems to see and reason about the compositional nature of visual scenes are fundamental to understanding our world. The complex relations between objects and their locations, ambiguities, and variations in the real-world environment can be better described in human language, naturally governed by grammatical rules and other modalities such as audio and depth. The models learned to bridge the gap between such modalities coupled with large-scale training data facilitate contextual reasoning, generalization, and prompt capabilities at test time. These models are referred to as foundational models. The output of such models can be modified through human-provided prompts without retraining, e.g., segmenting a particular object by providing a bounding box, having interactive dialogues by asking questions about an image or video scene or manipulating the robot's behavior through language instructions. In this survey, we provide a comprehensive review of such emerging foundational models, including typical architecture designs to combine different modalities (vision, text, audio, etc), training objectives (contrastive, generative), pre-training datasets, fine-tuning mechanisms, and the common prompting patterns; textual, visual, and heterogeneous. We discuss the open challenges and research directions for foundational models in computer vision, including difficulties in their evaluations and benchmarking, gaps in their real-world understanding, limitations of their contextual understanding, biases, vulnerability to adversarial attacks, and interpretability issues. We review recent developments in this field, covering a wide range of applications of foundation models systematically and comprehensively. A comprehensive list of foundational models studied in this work is available at \url{https://github.com/awaisrauf/Awesome-CV-Foundational-Models}.

Syed Talal Wasim, Muzammal Naseer, Salman Khan et al.

Adopting contrastive image-text pretrained models like CLIP towards video classification has gained attention due to its cost-effectiveness and competitive performance. However, recent works in this area face a trade-off. Finetuning the pretrained model to achieve strong supervised performance results in low zero-shot generalization. Similarly, freezing the backbone to retain zero-shot capability causes significant drop in supervised accuracy. Because of this, recent works in literature typically train separate models for supervised and zero-shot action recognition. In this work, we propose a multimodal prompt learning scheme that works to balance the supervised and zero-shot performance under a single unified training. Our prompting approach on the vision side caters for three aspects: 1) Global video-level prompts to model the data distribution; 2) Local frame-level prompts to provide per-frame discriminative conditioning; and 3) a summary prompt to extract a condensed video representation. Additionally, we define a prompting scheme on the text side to augment the textual context. Through this prompting scheme, we can achieve state-of-the-art zero-shot performance on Kinetics-600, HMDB51 and UCF101 while remaining competitive in the supervised setting. By keeping the pretrained backbone frozen, we optimize a much lower number of parameters and retain the existing general representation which helps achieve the strong zero-shot performance. Our codes/models are released at https://github.com/TalalWasim/Vita-CLIP.

Ishan Rajendrakumar Dave, Chen Chen, Mubarak Shah

Visual private information leakage is an emerging key issue for the fast growing applications of video understanding like activity recognition. Existing approaches for mitigating privacy leakage in action recognition require privacy labels along with the action labels from the video dataset. However, annotating frames of video dataset for privacy labels is not feasible. Recent developments of self-supervised learning (SSL) have unleashed the untapped potential of the unlabeled data. For the first time, we present a novel training framework which removes privacy information from input video in a self-supervised manner without requiring privacy labels. Our training framework consists of three main components: anonymization function, self-supervised privacy removal branch, and action recognition branch. We train our framework using a minimax optimization strategy to minimize the action recognition cost function and maximize the privacy cost function through a contrastive self-supervised loss. Employing existing protocols of known-action and privacy attributes, our framework achieves a competitive action-privacy trade-off to the existing state-of-the-art supervised methods. In addition, we introduce a new protocol to evaluate the generalization of learned the anonymization function to novel-action and privacy attributes and show that our self-supervised framework outperforms existing supervised methods. Code available at: https://github.com/DAVEISHAN/SPAct

Jiale Cao, Yanwei Pang, Rao Muhammad Anwer et al.

We propose a novel one-step transformer-based person search framework, PSTR, that jointly performs person detection and re-identification (re-id) in a single architecture. PSTR comprises a person search-specialized (PSS) module that contains a detection encoder-decoder for person detection along with a discriminative re-id decoder for person re-id. The discriminative re-id decoder utilizes a multi-level supervision scheme with a shared decoder for discriminative re-id feature learning and also comprises a part attention block to encode relationship between different parts of a person. We further introduce a simple multi-scale scheme to support re-id across person instances at different scales. PSTR jointly achieves the diverse objectives of object-level recognition (detection) and instance-level matching (re-id). To the best of our knowledge, we are the first to propose an end-to-end one-step transformer-based person search framework. Experiments are performed on two popular benchmarks: CUHK-SYSU and PRW. Our extensive ablations reveal the merits of the proposed contributions. Further, the proposed PSTR sets a new state-of-the-art on both benchmarks. On the challenging PRW benchmark, PSTR achieves a mean average precision (mAP) score of 56.5%. The source code is available at \url{https://github.com/JialeCao001/PSTR}.

Swetha Sirnam, Mamshad Nayeem Rizve, Nina Shvetsova et al. · ibm-research, mit

Self-supervised learning on large-scale multi-modal datasets allows learning semantically meaningful embeddings in a joint multi-modal representation space without relying on human annotations. These joint embeddings enable zero-shot cross-modal tasks like retrieval and classification. However, these methods often struggle to generalize well on out-of-domain data as they ignore the semantic structure present in modality-specific embeddings. In this context, we propose a novel Semantic-Structure-Preserving Consistency approach to improve generalizability by preserving the modality-specific relationships in the joint embedding space. To capture modality-specific semantic relationships between samples, we propose to learn multiple anchors and represent the multifaceted relationship between samples with respect to their relationship with these anchors. To assign multiple anchors to each sample, we propose a novel Multi-Assignment Sinkhorn-Knopp algorithm. Our experimentation demonstrates that our proposed approach learns semantically meaningful anchors in a self-supervised manner. Furthermore, our evaluation on MSR-VTT and YouCook2 datasets demonstrates that our proposed multi-anchor assignment based solution achieves state-of-the-art performance and generalizes to both inand out-of-domain datasets. Code: https://github.com/Swetha5/Multi_Sinkhorn_Knopp

Nicolae-Catalin Ristea, Florinel-Alin Croitoru, Radu Tudor Ionescu et al.

We propose an efficient abnormal event detection model based on a lightweight masked auto-encoder (AE) applied at the video frame level. The novelty of the proposed model is threefold. First, we introduce an approach to weight tokens based on motion gradients, thus shifting the focus from the static background scene to the foreground objects. Second, we integrate a teacher decoder and a student decoder into our architecture, leveraging the discrepancy between the outputs given by the two decoders to improve anomaly detection. Third, we generate synthetic abnormal events to augment the training videos, and task the masked AE model to jointly reconstruct the original frames (without anomalies) and the corresponding pixel-level anomaly maps. Our design leads to an efficient and effective model, as demonstrated by the extensive experiments carried out on four benchmarks: Avenue, ShanghaiTech, UBnormal and UCSD Ped2. The empirical results show that our model achieves an excellent trade-off between speed and accuracy, obtaining competitive AUC scores, while processing 1655 FPS. Hence, our model is between 8 and 70 times faster than competing methods. We also conduct an ablation study to justify our design. Our code is freely available at: https://github.com/ristea/aed-mae.

Sijie Zhu, Linjie Yang, Chen Chen et al.

Visual Place Recognition (VPR) estimates the location of query images by matching them with images in a reference database. Conventional methods generally adopt aggregated CNN features for global retrieval and RANSAC-based geometric verification for reranking. However, RANSAC only employs geometric information but ignores other possible information that could be useful for reranking, e.g. local feature correlations, and attention values. In this paper, we propose a unified place recognition framework that handles both retrieval and reranking with a novel transformer model, named $R^{2}$Former. The proposed reranking module takes feature correlation, attention value, and xy coordinates into account, and learns to determine whether the image pair is from the same location. The whole pipeline is end-to-end trainable and the reranking module alone can also be adopted on other CNN or transformer backbones as a generic component. Remarkably, $R^{2}$Former significantly outperforms state-of-the-art methods on major VPR datasets with much less inference time and memory consumption. It also achieves the state-of-the-art on the hold-out MSLS challenge set and could serve as a simple yet strong solution for real-world large-scale applications. Experiments also show vision transformer tokens are comparable and sometimes better than CNN local features on local matching. The code is released at https://github.com/Jeff-Zilence/R2Former.

Syed Talal Wasim, Muhammad Uzair Khattak, Muzammal Naseer et al.

Recent video recognition models utilize Transformer models for long-range spatio-temporal context modeling. Video transformer designs are based on self-attention that can model global context at a high computational cost. In comparison, convolutional designs for videos offer an efficient alternative but lack long-range dependency modeling. Towards achieving the best of both designs, this work proposes Video-FocalNet, an effective and efficient architecture for video recognition that models both local and global contexts. Video-FocalNet is based on a spatio-temporal focal modulation architecture that reverses the interaction and aggregation steps of self-attention for better efficiency. Further, the aggregation step and the interaction step are both implemented using efficient convolution and element-wise multiplication operations that are computationally less expensive than their self-attention counterparts on video representations. We extensively explore the design space of focal modulation-based spatio-temporal context modeling and demonstrate our parallel spatial and temporal encoding design to be the optimal choice. Video-FocalNets perform favorably well against the state-of-the-art transformer-based models for video recognition on five large-scale datasets (Kinetics-400, Kinetics-600, SS-v2, Diving-48, and ActivityNet-1.3) at a lower computational cost. Our code/models are released at https://github.com/TalalWasim/Video-FocalNets.

Shruti Vyas, Chen Chen, Mubarak Shah

The existing work in cross-view geo-localization is based on images where a ground panorama is matched to an aerial image. In this work, we focus on ground videos instead of images which provides additional contextual cues which are important for this task. There are no existing datasets for this problem, therefore we propose GAMa dataset, a large-scale dataset with ground videos and corresponding aerial images. We also propose a novel approach to solve this problem. At clip-level, a short video clip is matched with corresponding aerial image and is later used to get video-level geo-localization of a long video. Moreover, we propose a hierarchical approach to further improve the clip-level geolocalization. It is a challenging dataset, unaligned and limited field of view, and our proposed method achieves a Top-1 recall rate of 19.4% and 45.1% @1.0mile. Code and dataset are available at following link: https://github.com/svyas23/GAMa.

Aisha Urooj Khan, Hilde Kuehne, Bo Wu et al. · ibm-research, mit

Answering questions about complex situations in videos requires not only capturing the presence of actors, objects, and their relations but also the evolution of these relationships over time. A situation hyper-graph is a representation that describes situations as scene sub-graphs for video frames and hyper-edges for connected sub-graphs and has been proposed to capture all such information in a compact structured form. In this work, we propose an architecture for Video Question Answering (VQA) that enables answering questions related to video content by predicting situation hyper-graphs, coined Situation Hyper-Graph based Video Question Answering (SHG-VQA). To this end, we train a situation hyper-graph decoder to implicitly identify graph representations with actions and object/human-object relationships from the input video clip. and to use cross-attention between the predicted situation hyper-graphs and the question embedding to predict the correct answer. The proposed method is trained in an end-to-end manner and optimized by a VQA loss with the cross-entropy function and a Hungarian matching loss for the situation graph prediction. The effectiveness of the proposed architecture is extensively evaluated on two challenging benchmarks: AGQA and STAR. Our results show that learning the underlying situation hyper-graphs helps the system to significantly improve its performance for novel challenges of video question-answering tasks.

Omkar Thawakar, Sanath Narayan, Hisham Cholakkal et al.

Existing video instance segmentation (VIS) approaches generally follow a closed-world assumption, where only seen category instances are identified and spatio-temporally segmented at inference. Open-world formulation relaxes the close-world static-learning assumption as follows: (a) first, it distinguishes a set of known categories as well as labels an unknown object as `unknown' and then (b) it incrementally learns the class of an unknown as and when the corresponding semantic labels become available. We propose the first open-world VIS approach, named OW-VISFormer, that introduces a novel feature enrichment mechanism and a spatio-temporal objectness (STO) module. The feature enrichment mechanism based on a light-weight auxiliary network aims at accurate pixel-level (unknown) object delineation from the background as well as distinguishing category-specific known semantic classes. The STO module strives to generate instance-level pseudo-labels by enhancing the foreground activations through a contrastive loss. Moreover, we also introduce an extensive experimental protocol to measure the characteristics of OW-VIS. Our OW-VISFormer performs favorably against a solid baseline in OW-VIS setting. Further, we evaluate our contributions in the standard fully-supervised VIS setting by integrating them into the recent SeqFormer, achieving an absolute gain of 1.6\% AP on Youtube-VIS 2019 val. set. Lastly, we show the generalizability of our contributions for the open-world detection (OWOD) setting, outperforming the best existing OWOD method in the literature. Code, models along with OW-VIS splits are available at \url{https://github.com/OmkarThawakar/OWVISFormer}.

Asif Hanif, Muzammal Naseer, Salman Khan et al.

It is imperative to ensure the robustness of deep learning models in critical applications such as, healthcare. While recent advances in deep learning have improved the performance of volumetric medical image segmentation models, these models cannot be deployed for real-world applications immediately due to their vulnerability to adversarial attacks. We present a 3D frequency domain adversarial attack for volumetric medical image segmentation models and demonstrate its advantages over conventional input or voxel domain attacks. Using our proposed attack, we introduce a novel frequency domain adversarial training approach for optimizing a robust model against voxel and frequency domain attacks. Moreover, we propose frequency consistency loss to regulate our frequency domain adversarial training that achieves a better tradeoff between model's performance on clean and adversarial samples. Code is publicly available at https://github.com/asif-hanif/vafa.

Omkar Thawakar, Rao Muhammad Anwer, Jorma Laaksonen et al.

Accurate 3D mitochondria instance segmentation in electron microscopy (EM) is a challenging problem and serves as a prerequisite to empirically analyze their distributions and morphology. Most existing approaches employ 3D convolutions to obtain representative features. However, these convolution-based approaches struggle to effectively capture long-range dependencies in the volume mitochondria data, due to their limited local receptive field. To address this, we propose a hybrid encoder-decoder framework based on a split spatio-temporal attention module that efficiently computes spatial and temporal self-attentions in parallel, which are later fused through a deformable convolution. Further, we introduce a semantic foreground-background adversarial loss during training that aids in delineating the region of mitochondria instances from the background clutter. Our extensive experiments on three benchmarks, Lucchi, MitoEM-R and MitoEM-H, reveal the benefits of the proposed contributions achieving state-of-the-art results on all three datasets. Our code and models are available at https://github.com/OmkarThawakar/STT-UNET.

Ishan Rajendrakumar Dave, Mamshad Nayeem Rizve, Chen Chen et al.

Semi-Supervised Learning can be more beneficial for the video domain compared to images because of its higher annotation cost and dimensionality. Besides, any video understanding task requires reasoning over both spatial and temporal dimensions. In order to learn both the static and motion related features for the semi-supervised action recognition task, existing methods rely on hard input inductive biases like using two-modalities (RGB and Optical-flow) or two-stream of different playback rates. Instead of utilizing unlabeled videos through diverse input streams, we rely on self-supervised video representations, particularly, we utilize temporally-invariant and temporally-distinctive representations. We observe that these representations complement each other depending on the nature of the action. Based on this observation, we propose a student-teacher semi-supervised learning framework, TimeBalance, where we distill the knowledge from a temporally-invariant and a temporally-distinctive teacher. Depending on the nature of the unlabeled video, we dynamically combine the knowledge of these two teachers based on a novel temporal similarity-based reweighting scheme. Our method achieves state-of-the-art performance on three action recognition benchmarks: UCF101, HMDB51, and Kinetics400. Code: https://github.com/DAVEISHAN/TimeBalance

Shehan Munasinghe, Rusiru Thushara, Muhammad Maaz et al.

Extending image-based Large Multimodal Models (LMMs) to videos is challenging due to the inherent complexity of video data. The recent approaches extending image-based LMMs to videos either lack the grounding capabilities (e.g., VideoChat, Video-ChatGPT, Video-LLaMA) or do not utilize the audio-signals for better video understanding (e.g., Video-ChatGPT). Addressing these gaps, we propose PG-Video-LLaVA, the first LMM with pixel-level grounding capability, integrating audio cues by transcribing them into text to enrich video-context understanding. Our framework uses an off-the-shelf tracker and a novel grounding module, enabling it to spatially localize objects in videos following user instructions. We evaluate PG-Video-LLaVA using video-based generative and question-answering benchmarks and introduce new benchmarks specifically designed to measure prompt-based object grounding performance in videos. Further, we propose the use of Vicuna over GPT-3.5, as utilized in Video-ChatGPT, for video-based conversation benchmarking, ensuring reproducibility of results which is a concern with the proprietary nature of GPT-3.5. Our framework builds on SoTA image-based LLaVA model and extends its advantages to the video domain, delivering promising gains on video-based conversation and grounding tasks. Project Page: https://github.com/mbzuai-oryx/Video-LLaVA

Antonio Barbalau, Radu Tudor Ionescu, Mariana-Iuliana Georgescu et al.

A self-supervised multi-task learning (SSMTL) framework for video anomaly detection was recently introduced in literature. Due to its highly accurate results, the method attracted the attention of many researchers. In this work, we revisit the self-supervised multi-task learning framework, proposing several updates to the original method. First, we study various detection methods, e.g. based on detecting high-motion regions using optical flow or background subtraction, since we believe the currently used pre-trained YOLOv3 is suboptimal, e.g. objects in motion or objects from unknown classes are never detected. Second, we modernize the 3D convolutional backbone by introducing multi-head self-attention modules, inspired by the recent success of vision transformers. As such, we alternatively introduce both 2D and 3D convolutional vision transformer (CvT) blocks. Third, in our attempt to further improve the model, we study additional self-supervised learning tasks, such as predicting segmentation maps through knowledge distillation, solving jigsaw puzzles, estimating body pose through knowledge distillation, predicting masked regions (inpainting), and adversarial learning with pseudo-anomalies. We conduct experiments to assess the performance impact of the introduced changes. Upon finding more promising configurations of the framework, dubbed SSMTL++v1 and SSMTL++v2, we extend our preliminary experiments to more data sets, demonstrating that our performance gains are consistent across all data sets. In most cases, our results on Avenue, ShanghaiTech and UBnormal raise the state-of-the-art performance bar to a new level.

Mitch Hill, Jonathan Mitchell, Chu Chen et al.

This work presents strategies to learn an Energy-Based Model (EBM) according to the desired length of its MCMC sampling trajectories. MCMC trajectories of different lengths correspond to models with different purposes. Our experiments cover three different trajectory magnitudes and learning outcomes: 1) shortrun sampling for image generation; 2) midrun sampling for classifier-agnostic adversarial defense; and 3) longrun sampling for principled modeling of image probability densities. To achieve these outcomes, we introduce three novel methods of MCMC initialization for negative samples used in Maximum Likelihood (ML) learning. With standard network architectures and an unaltered ML objective, our MCMC initialization methods alone enable significant performance gains across the three applications that we investigate. Our results include state-of-the-art FID scores for unnormalized image densities on the CIFAR-10 and ImageNet datasets; state-of-the-art adversarial defense on CIFAR-10 among purification methods and the first EBM defense on ImageNet; and scalable techniques for learning valid probability densities. Code for this project can be found at https://github.com/point0bar1/ebm-life-cycle.

Mamshad Nayeem Rizve, Navid Kardan, Salman Khan et al.

Semi-supervised learning (SSL) is one of the dominant approaches to address the annotation bottleneck of supervised learning. Recent SSL methods can effectively leverage a large repository of unlabeled data to improve performance while relying on a small set of labeled data. One common assumption in most SSL methods is that the labeled and unlabeled data are from the same data distribution. However, this is hardly the case in many real-world scenarios, which limits their applicability. In this work, instead, we attempt to solve the challenging open-world SSL problem that does not make such an assumption. In the open-world SSL problem, the objective is to recognize samples of known classes, and simultaneously detect and cluster samples belonging to novel classes present in unlabeled data. This work introduces OpenLDN that utilizes a pairwise similarity loss to discover novel classes. Using a bi-level optimization rule this pairwise similarity loss exploits the information available in the labeled set to implicitly cluster novel class samples, while simultaneously recognizing samples from known classes. After discovering novel classes, OpenLDN transforms the open-world SSL problem into a standard SSL problem to achieve additional performance gains using existing SSL methods. Our extensive experiments demonstrate that OpenLDN outperforms the current state-of-the-art methods on multiple popular classification benchmarks while providing a better accuracy/training time trade-off.

Brandon Clark, Alec Kerrigan, Parth Parag Kulkarni et al.

Determining the exact latitude and longitude that a photo was taken is a useful and widely applicable task, yet it remains exceptionally difficult despite the accelerated progress of other computer vision tasks. Most previous approaches have opted to learn a single representation of query images, which are then classified at different levels of geographic granularity. These approaches fail to exploit the different visual cues that give context to different hierarchies, such as the country, state, and city level. To this end, we introduce an end-to-end transformer-based architecture that exploits the relationship between different geographic levels (which we refer to as hierarchies) and the corresponding visual scene information in an image through hierarchical cross-attention. We achieve this by learning a query for each geographic hierarchy and scene type. Furthermore, we learn a separate representation for different environmental scenes, as different scenes in the same location are often defined by completely different visual features. We achieve state of the art street level accuracy on 4 standard geo-localization datasets : Im2GPS, Im2GPS3k, YFCC4k, and YFCC26k, as well as qualitatively demonstrate how our method learns different representations for different visual hierarchies and scenes, which has not been demonstrated in the previous methods. These previous testing datasets mostly consist of iconic landmarks or images taken from social media, which makes them either a memorization task, or biased towards certain places. To address this issue we introduce a much harder testing dataset, Google-World-Streets-15k, comprised of images taken from Google Streetview covering the whole planet and present state of the art results. Our code will be made available in the camera-ready version.

Nazmul Karim, Mamshad Nayeem Rizve, Nazanin Rahnavard et al.

Supervised deep learning methods require a large repository of annotated data; hence, label noise is inevitable. Training with such noisy data negatively impacts the generalization performance of deep neural networks. To combat label noise, recent state-of-the-art methods employ some sort of sample selection mechanism to select a possibly clean subset of data. Next, an off-the-shelf semi-supervised learning method is used for training where rejected samples are treated as unlabeled data. Our comprehensive analysis shows that current selection methods disproportionately select samples from easy (fast learnable) classes while rejecting those from relatively harder ones. This creates class imbalance in the selected clean set and in turn, deteriorates performance under high label noise. In this work, we propose UNICON, a simple yet effective sample selection method which is robust to high label noise. To address the disproportionate selection of easy and hard samples, we introduce a Jensen-Shannon divergence based uniform selection mechanism which does not require any probabilistic modeling and hyperparameter tuning. We complement our selection method with contrastive learning to further combat the memorization of noisy labels. Extensive experimentation on multiple benchmark datasets demonstrates the effectiveness of UNICON; we obtain an 11.4% improvement over the current state-of-the-art on CIFAR100 dataset with a 90% noise rate. Our code is publicly available

Sarinda Samarasinghe, Mamshad Nayeem Rizve, Navid Kardan et al.

Few-shot video action recognition is an effective approach to recognizing new categories with only a few labeled examples, thereby reducing the challenges associated with collecting and annotating large-scale video datasets. Existing methods in video action recognition rely on large labeled datasets from the same domain. However, this setup is not realistic as novel categories may come from different data domains that may have different spatial and temporal characteristics. This dissimilarity between the source and target domains can pose a significant challenge, rendering traditional few-shot action recognition techniques ineffective. To address this issue, in this work, we propose a novel cross-domain few-shot video action recognition method that leverages self-supervised learning and curriculum learning to balance the information from the source and target domains. To be particular, our method employs a masked autoencoder-based self-supervised training objective to learn from both source and target data in a self-supervised manner. Then a progressive curriculum balances learning the discriminative information from the source dataset with the generic information learned from the target domain. Initially, our curriculum utilizes supervised learning to learn class discriminative features from the source data. As the training progresses, we transition to learning target-domain-specific features. We propose a progressive curriculum to encourage the emergence of rich features in the target domain based on class discriminative supervised features in the source domain. We evaluate our method on several challenging benchmark datasets and demonstrate that our approach outperforms existing cross-domain few-shot learning techniques. Our code is available at https://github.com/Sarinda251/CDFSL-V

Tushar Sangam, Ishan Rajendrakumar Dave, Waqas Sultani et al.

Drone-to-drone detection using visual feed has crucial applications, such as detecting drone collisions, detecting drone attacks, or coordinating flight with other drones. However, existing methods are computationally costly, follow non-end-to-end optimization, and have complex multi-stage pipelines, making them less suitable for real-time deployment on edge devices. In this work, we propose a simple yet effective framework, \textit{TransVisDrone}, that provides an end-to-end solution with higher computational efficiency. We utilize CSPDarkNet-53 network to learn object-related spatial features and VideoSwin model to improve drone detection in challenging scenarios by learning spatio-temporal dependencies of drone motion. Our method achieves state-of-the-art performance on three challenging real-world datasets (Average Precision@0.5IOU): NPS 0.95, FLDrones 0.75, and AOT 0.80, and a higher throughput than previous methods. We also demonstrate its deployment capability on edge devices and its usefulness in detecting drone-collision (encounter). Project: \url{https://tusharsangam.github.io/TransVisDrone-project-page/}.

Tristan de Blegiers, Ishan Rajendrakumar Dave, Adeel Yousaf et al.

Recognizing and comprehending human actions and gestures is a crucial perception requirement for robots to interact with humans and carry out tasks in diverse domains, including service robotics, healthcare, and manufacturing. Event cameras, with their ability to capture fast-moving objects at a high temporal resolution, offer new opportunities compared to standard action recognition in RGB videos. However, previous research on event camera action recognition has primarily focused on sensor-specific network architectures and image encoding, which may not be suitable for new sensors and limit the use of recent advancements in transformer-based architectures. In this study, we employ a computationally efficient model, namely the video transformer network (VTN), which initially acquires spatial embeddings per event-frame and then utilizes a temporal self-attention mechanism. In order to better adopt the VTN for the sparse and fine-grained nature of event data, we design Event-Contrastive Loss ($\mathcal{L}_{EC}$) and event-specific augmentations. Proposed $\mathcal{L}_{EC}$ promotes learning fine-grained spatial cues in the spatial backbone of VTN by contrasting temporally misaligned frames. We evaluate our method on real-world action recognition of N-EPIC Kitchens dataset, and achieve state-of-the-art results on both protocols - testing in seen kitchen (\textbf{74.9\%} accuracy) and testing in unseen kitchens (\textbf{42.43\% and 46.66\% Accuracy}). Our approach also takes less computation time compared to competitive prior approaches, which demonstrates the potential of our framework \textit{EventTransAct} for real-world applications of event-camera based action recognition. Project Page: \url{https://tristandb8.github.io/EventTransAct_webpage/}

Jyoti Kini, Ajmal Mian, Mubarak Shah

We propose a method for joint detection and tracking of multiple objects in 3D point clouds, a task conventionally treated as a two-step process comprising object detection followed by data association. Our method embeds both steps into a single end-to-end trainable network eliminating the dependency on external object detectors. Our model exploits temporal information employing multiple frames to detect objects and track them in a single network, thereby making it a utilitarian formulation for real-world scenarios. Computing affinity matrix by employing features similarity across consecutive point cloud scans forms an integral part of visual tracking. We propose an attention-based refinement module to refine the affinity matrix by suppressing erroneous correspondences. The module is designed to capture the global context in affinity matrix by employing self-attention within each affinity matrix and cross-attention across a pair of affinity matrices. Unlike competing approaches, our network does not require complex post-processing algorithms, and processes raw LiDAR frames to directly output tracking results. We demonstrate the effectiveness of our method on the three tracking benchmarks: JRDB, Waymo, and KITTI. Experimental evaluations indicate the ability of our model to generalize well across datasets.

Rajat Modi, Aayush Jung Rana, Akash Kumar et al.

Beyond possessing large enough size to feed data hungry machines (eg, transformers), what attributes measure the quality of a dataset? Assuming that the definitions of such attributes do exist, how do we quantify among their relative existences? Our work attempts to explore these questions for video action detection. The task aims to spatio-temporally localize an actor and assign a relevant action class. We first analyze the existing datasets on video action detection and discuss their limitations. Next, we propose a new dataset, Multi Actor Multi Action (MAMA) which overcomes these limitations and is more suitable for real world applications. In addition, we perform a biasness study which analyzes a key property differentiating videos from static images: the temporal aspect. This reveals if the actions in these datasets really need the motion information of an actor, or whether they predict the occurrence of an action even by looking at a single frame. Finally, we investigate the widely held assumptions on the importance of temporal ordering: is temporal ordering important for detecting these actions? Such extreme experiments show existence of biases which have managed to creep into existing methods inspite of careful modeling.

Madeline C. Schiappa, Yogesh S. Rawat, Mubarak Shah

The remarkable success of deep learning in various domains relies on the availability of large-scale annotated datasets. However, obtaining annotations is expensive and requires great effort, which is especially challenging for videos. Moreover, the use of human-generated annotations leads to models with biased learning and poor domain generalization and robustness. As an alternative, self-supervised learning provides a way for representation learning which does not require annotations and has shown promise in both image and video domains. Different from the image domain, learning video representations are more challenging due to the temporal dimension, bringing in motion and other environmental dynamics. This also provides opportunities for video-exclusive ideas that advance self-supervised learning in the video and multimodal domain. In this survey, we provide a review of existing approaches on self-supervised learning focusing on the video domain. We summarize these methods into four different categories based on their learning objectives: 1) pretext tasks, 2) generative learning, 3) contrastive learning, and 4) cross-modal agreement. We further introduce the commonly used datasets, downstream evaluation tasks, insights into the limitations of existing works, and the potential future directions in this area.

Rohit Gupta, Naveed Akhtar, Ajmal Mian et al.

Contrastive self-supervised learning (CSL) has managed to match or surpass the performance of supervised learning in image and video classification. However, it is still largely unknown if the nature of the representations induced by the two learning paradigms is similar. We investigate this under the lens of adversarial robustness. Our analysis of the problem reveals that CSL has intrinsically higher sensitivity to perturbations over supervised learning. We identify the uniform distribution of data representation over a unit hypersphere in the CSL representation space as the key contributor to this phenomenon. We establish that this is a result of the presence of false negative pairs in the training process, which increases model sensitivity to input perturbations. Our finding is supported by extensive experiments for image and video classification using adversarial perturbations and other input corruptions. We devise a strategy to detect and remove false negative pairs that is simple, yet effective in improving model robustness with CSL training. We close up to 68% of the robustness gap between CSL and its supervised counterpart. Finally, we contribute to adversarial learning by incorporating our method in CSL. We demonstrate an average gain of about 5% over two different state-of-the-art methods in this domain.

Jyoti Kini, Fahad Shahbaz Khan, Salman Khan et al.

We propose a novel self-supervised Video Object Segmentation (VOS) approach that strives to achieve better object-background discriminability for accurate object segmentation. Distinct from previous self-supervised VOS methods, our approach is based on a discriminative learning loss formulation that takes into account both object and background information to ensure object-background discriminability, rather than using only object appearance. The discriminative learning loss comprises cutout-based reconstruction (cutout region represents part of a frame, whose pixels are replaced with some constant values) and tag prediction loss terms. The cutout-based reconstruction term utilizes a simple cutout scheme to learn the pixel-wise correspondence between the current and previous frames in order to reconstruct the original current frame with added cutout region in it. The introduced cutout patch guides the model to focus as much on the significant features of the object of interest as the less significant ones, thereby implicitly equipping the model to address occlusion-based scenarios. Next, the tag prediction term encourages object-background separability by grouping tags of all pixels in the cutout region that are similar, while separating them from the tags of the rest of the reconstructed frame pixels. Additionally, we introduce a zoom-in scheme that addresses the problem of small object segmentation by capturing fine structural information at multiple scales. Our proposed approach, termed CT-VOS, achieves state-of-the-art results on two challenging benchmarks: DAVIS-2017 and Youtube-VOS. A detailed ablation showcases the importance of the proposed loss formulation to effectively capture object-background discriminability and the impact of our zoom-in scheme to accurately segment small-sized objects.

Mamshad Nayeem Rizve, Navid Kardan, Mubarak Shah

Deep learning is pushing the state-of-the-art in many computer vision applications. However, it relies on large annotated data repositories, and capturing the unconstrained nature of the real-world data is yet to be solved. Semi-supervised learning (SSL) complements the annotated training data with a large corpus of unlabeled data to reduce annotation cost. The standard SSL approach assumes unlabeled data are from the same distribution as annotated data. Recently, a more realistic SSL problem, called open-world SSL, is introduced, where the unannotated data might contain samples from unknown classes. In this paper, we propose a novel pseudo-label based approach to tackle SSL in open-world setting. At the core of our method, we utilize sample uncertainty and incorporate prior knowledge about class distribution to generate reliable class-distribution-aware pseudo-labels for unlabeled data belonging to both known and unknown classes. Our extensive experimentation showcases the effectiveness of our approach on several benchmark datasets, where it substantially outperforms the existing state-of-the-art on seven diverse datasets including CIFAR-100 (~17%), ImageNet-100 (~5%), and Tiny ImageNet (~9%). We also highlight the flexibility of our approach in solving novel class discovery task, demonstrate its stability in dealing with imbalanced data, and complement our approach with a technique to estimate the number of novel classes

Vicente Vivanco Cepeda, Gaurav Kumar Nayak, Mubarak Shah

Worldwide Geo-localization aims to pinpoint the precise location of images taken anywhere on Earth. This task has considerable challenges due to immense variation in geographic landscapes. The image-to-image retrieval-based approaches fail to solve this problem on a global scale as it is not feasible to construct a large gallery of images covering the entire world. Instead, existing approaches divide the globe into discrete geographic cells, transforming the problem into a classification task. However, their performance is limited by the predefined classes and often results in inaccurate localizations when an image's location significantly deviates from its class center. To overcome these limitations, we propose GeoCLIP, a novel CLIP-inspired Image-to-GPS retrieval approach that enforces alignment between the image and its corresponding GPS locations. GeoCLIP's location encoder models the Earth as a continuous function by employing positional encoding through random Fourier features and constructing a hierarchical representation that captures information at varying resolutions to yield a semantically rich high-dimensional feature suitable to use even beyond geo-localization. To the best of our knowledge, this is the first work employing GPS encoding for geo-localization. We demonstrate the efficacy of our method via extensive experiments and ablations on benchmark datasets. We achieve competitive performance with just 20% of training data, highlighting its effectiveness even in limited-data settings. Furthermore, we qualitatively demonstrate geo-localization using a text query by leveraging CLIP backbone of our image encoder. The project webpage is available at: https://vicentevivan.github.io/GeoCLIP

Jyoti Kini, Mubarak Shah

Video Instance Segmentation is a fundamental computer vision task that deals with segmenting and tracking object instances across a video sequence. Most existing methods typically accomplish this task by employing a multi-stage top-down approach that usually involves separate networks to detect and segment objects in each frame, followed by associating these detections in consecutive frames using a learned tracking head. In this work, however, we introduce a simple end-to-end trainable bottom-up approach to achieve instance mask predictions at the pixel-level granularity, instead of the typical region-proposals-based approach. Unlike contemporary frame-based models, our network pipeline processes an input video clip as a single 3D volume to incorporate temporal information. The central idea of our formulation is to solve the video instance segmentation task as a tag assignment problem, such that generating distinct tag values essentially separates individual object instances across the video sequence (here each tag could be any arbitrary value between 0 and 1). To this end, we propose a novel spatio-temporal tagging loss that allows for sufficient separation of different objects as well as necessary identification of different instances of the same object. Furthermore, we present a tag-based attention module that improves instance tags, while concurrently learning instance propagation within a video. Evaluations demonstrate that our method provides competitive results on YouTube-VIS and DAVIS-19 datasets, and has minimum run-time compared to other state-of-the-art performance methods.

Joseph Fioresi, Ishan Rajendrakumar Dave, Mubarak Shah

Video anomaly detection (VAD) without human monitoring is a complex computer vision task that can have a positive impact on society if implemented successfully. While recent advances have made significant progress in solving this task, most existing approaches overlook a critical real-world concern: privacy. With the increasing popularity of artificial intelligence technologies, it becomes crucial to implement proper AI ethics into their development. Privacy leakage in VAD allows models to pick up and amplify unnecessary biases related to people's personal information, which may lead to undesirable decision making. In this paper, we propose TeD-SPAD, a privacy-aware video anomaly detection framework that destroys visual private information in a self-supervised manner. In particular, we propose the use of a temporally-distinct triplet loss to promote temporally discriminative features, which complements current weakly-supervised VAD methods. Using TeD-SPAD, we achieve a positive trade-off between privacy protection and utility anomaly detection performance on three popular weakly supervised VAD datasets: UCF-Crime, XD-Violence, and ShanghaiTech. Our proposed anonymization model reduces private attribute prediction by 32.25% while only reducing frame-level ROC AUC on the UCF-Crime anomaly detection dataset by 3.69%. Project Page: https://joefioresi718.github.io/TeD-SPAD_webpage/

Sirnam Swetha, Jinyu Yang, Tal Neiman et al.

Recent advancements in Multimodal Large Language Models (MLLMs) have revolutionized the field of vision-language understanding by integrating visual perception capabilities into Large Language Models (LLMs). The prevailing trend in this field involves the utilization of a vision encoder derived from vision-language contrastive learning (CL), showing expertise in capturing overall representations while facing difficulties in capturing detailed local patterns. In this work, we focus on enhancing the visual representations for MLLMs by combining high-frequency and detailed visual representations, obtained through masked image modeling (MIM), with semantically-enriched low-frequency representations captured by CL. To achieve this goal, we introduce X-Former which is a lightweight transformer module designed to exploit the complementary strengths of CL and MIM through an innovative interaction mechanism. Specifically, X-Former first bootstraps vision-language representation learning and multimodal-to-multimodal generative learning from two frozen vision encoders, i.e., CLIP-ViT (CL-based) and MAE-ViT (MIM-based). It further bootstraps vision-to-language generative learning from a frozen LLM to ensure visual features from X-Former can be interpreted by the LLM. To demonstrate the effectiveness of our approach, we assess its performance on tasks demanding detailed visual understanding. Extensive evaluations indicate that X-Former excels in visual reasoning tasks involving both structural and semantic categories in the GQA dataset. Assessment on fine-grained visual perception benchmark further confirms its superior capabilities in visual understanding.

Rohit Gupta, Naveed Akhtar, Gaurav Kumar Nayak et al.

Black-box adversarial attacks present a realistic threat to action recognition systems. Existing black-box attacks follow either a query-based approach where an attack is optimized by querying the target model, or a transfer-based approach where attacks are generated using a substitute model. While these methods can achieve decent fooling rates, the former tends to be highly query-inefficient while the latter assumes extensive knowledge of the black-box model's training data. In this paper, we propose a new attack on action recognition that addresses these shortcomings by generating perturbations to disrupt the features learned by a pre-trained substitute model to reduce the number of queries. By using a nearly disjoint dataset to train the substitute model, our method removes the requirement that the substitute model be trained using the same dataset as the target model, and leverages queries to the target model to retain the fooling rate benefits provided by query-based methods. This ultimately results in attacks which are more transferable than conventional black-box attacks. Through extensive experiments, we demonstrate highly query-efficient black-box attacks with the proposed framework. Our method achieves 8% and 12% higher deception rates compared to state-of-the-art query-based and transfer-based attacks, respectively.

Manu S Pillai, Mamshad Nayeem Rizve, Mubarak Shah

Cross-view video geo-localization (CVGL) aims to derive GPS trajectories from street-view videos by aligning them with aerial-view images. Despite their promising performance, current CVGL methods face significant challenges. These methods use camera and odometry data, typically absent in real-world scenarios. They utilize multiple adjacent frames and various encoders for feature extraction, resulting in high computational costs. Moreover, these approaches independently predict each street-view frame's location, resulting in temporally inconsistent GPS trajectories. To address these challenges, in this work, we propose GAReT, a fully transformer-based method for CVGL that does not require camera and odometry data. We introduce GeoAdapter, a transformer-adapter module designed to efficiently aggregate image-level representations and adapt them for video inputs. Specifically, we train a transformer encoder on video frames and aerial images, then freeze the encoder to optimize the GeoAdapter module to obtain video-level representation. To address temporally inconsistent trajectories, we introduce TransRetriever, an encoder-decoder transformer model that predicts GPS locations of street-view frames by encoding top-k nearest neighbor predictions per frame and auto-regressively decoding the best neighbor based on the previous frame's predictions. Our method's effectiveness is validated through extensive experiments, demonstrating state-of-the-art performance on benchmark datasets. Our code is available at https://github.com/manupillai308/GAReT.

Fabio De Sousa Ribeiro, Kevin Duarte, Miles Everett et al.

Capsule networks were proposed as an alternative approach to Convolutional Neural Networks (CNNs) for learning object-centric representations, which can be leveraged for improved generalization and sample complexity. Unlike CNNs, capsule networks are designed to explicitly model part-whole hierarchical relationships by using groups of neurons to encode visual entities, and learn the relationships between those entities. Promising early results achieved by capsule networks have motivated the deep learning community to continue trying to improve their performance and scalability across several application areas. However, a major hurdle for capsule network research has been the lack of a reliable point of reference for understanding their foundational ideas and motivations. The aim of this survey is to provide a comprehensive overview of the capsule network research landscape, which will serve as a valuable resource for the community going forward. To that end, we start with an introduction to the fundamental concepts and motivations behind capsule networks, such as equivariant inference in computer vision. We then cover the technical advances in the capsule routing mechanisms and the various formulations of capsule networks, e.g. generative and geometric. Additionally, we provide a detailed explanation of how capsule networks relate to the popular attention mechanism in Transformers, and highlight non-trivial conceptual similarities between them in the context of representation learning. Afterwards, we explore the extensive applications of capsule networks in computer vision, video and motion, graph representation learning, natural language processing, medical imaging and many others. To conclude, we provide an in-depth discussion regarding the main hurdles in capsule network research, and highlight promising research directions for future work.

Guo-Jun Qi, Mubarak Shah

In this paper, we review adversarial pretraining of self-supervised deep networks including both convolutional neural networks and vision transformers. Unlike the adversarial training with access to labeled examples, adversarial pretraining is complicated as it only has access to unlabeled examples. To incorporate adversaries into pretraining models on either input or feature level, we find that existing approaches are largely categorized into two groups: memory-free instance-wise attacks imposing worst-case perturbations on individual examples, and memory-based adversaries shared across examples over iterations. In particular, we review several representative adversarial pretraining models based on Contrastive Learning (CL) and Masked Image Modeling (MIM), respectively, two popular self-supervised pretraining methods in literature. We also review miscellaneous issues about computing overheads, input-/feature-level adversaries, as well as other adversarial pretraining approaches beyond the above two groups. Finally, we discuss emerging trends and future directions about the relations between adversarial and cooperative pretraining, unifying adversarial CL and MIM pretraining, and the trade-off between accuracy and robustness in adversarial pretraining.

Weitai Kang, Gaowen Liu, Mubarak Shah et al.

Different from Object Detection, Visual Grounding deals with detecting a bounding box for each text-image pair. This one box for each text-image data provides sparse supervision signals. Although previous works achieve impressive results, their passive utilization of annotation, i.e. the sole use of the box annotation as regression ground truth, results in a suboptimal performance. In this paper, we present SegVG, a novel method transfers the box-level annotation as Segmentation signals to provide an additional pixel-level supervision for Visual Grounding. Specifically, we propose the Multi-layer Multi-task Encoder-Decoder as the target grounding stage, where we learn a regression query and multiple segmentation queries to ground the target by regression and segmentation of the box in each decoding layer, respectively. This approach allows us to iteratively exploit the annotation as signals for both box-level regression and pixel-level segmentation. Moreover, as the backbones are typically initialized by pretrained parameters learned from unimodal tasks and the queries for both regression and segmentation are static learnable embeddings, a domain discrepancy remains among these three types of features, which impairs subsequent target grounding. To mitigate this discrepancy, we introduce the Triple Alignment module, where the query, text, and vision tokens are triangularly updated to share the same space by triple attention mechanism. Extensive experiments on five widely used datasets validate our state-of-the-art (SOTA) performance.

Peiyu Yang, Naveed Akhtar, Mubarak Shah et al.

Trustworthy machine learning necessitates meticulous regulation of model reliance on non-robust features. We propose a framework to delineate and regulate such features by attributing model predictions to the input. Within our approach, robust feature attributions exhibit a certain consistency, while non-robust feature attributions are susceptible to fluctuations. This behavior allows identification of correlation between model reliance on non-robust features and smoothness of marginal density of the input samples. Hence, we uniquely regularize the gradients of the marginal density w.r.t. the input features for robustness. We also devise an efficient implementation of our regularization to address the potential numerical instability of the underlying optimization process. Moreover, we analytically reveal that, as opposed to our marginal density smoothing, the prevalent input gradient regularization smoothens conditional or joint density of the input, which can cause limited robustness. Our experiments validate the effectiveness of the proposed method, providing clear evidence of its capability to address the feature leakage problem and mitigate spurious correlations. Extensive results further establish that our technique enables the model to exhibit robustness against perturbations in pixel values, input gradients, and density.

Weitai Kang, Haifeng Huang, Yuzhang Shang et al.

Recent advancements in 3D Large Language Models (3DLLMs) have highlighted their potential in building general-purpose agents in the 3D real world, yet challenges remain due to the lack of high-quality robust instruction-following data, leading to limited discriminative power and generalization of 3DLLMs. In this paper, we introduce Robin3D, a powerful 3DLLM trained on large-scale instruction-following data generated by our novel data engine, Robust Instruction Generation (RIG) engine. RIG generates two key instruction data: 1) the Adversarial Instruction-following data, which features mixed negative and positive samples to enhance the model's discriminative understanding. 2) the Diverse Instruction-following data, which contains various instruction styles to enhance model's generalization. As a result, we construct 1 million instruction-following data, consisting of 344K Adversarial samples, 508K Diverse samples, and 165K benchmark training set samples. To better handle these complex instructions, Robin3D first incorporates Relation-Augmented Projector to enhance spatial understanding, and then strengthens the object referring and grounding ability through ID-Feature Bonding. Robin3D consistently outperforms previous methods across five widely-used 3D multimodal learning benchmarks, without the need for task-specific fine-tuning. Notably, we achieve a 7.8\% improvement in the grounding task (Multi3DRefer) and a 6.9\% improvement in the captioning task (Scan2Cap).

Adeel Yousaf, Muzammal Naseer, Salman Khan et al.

Vision-language models (VLMs) classify the query video by calculating a similarity score between the visual features and text-based class label representations. Recently, large language models (LLMs) have been used to enrich the text-based class labels by enhancing the descriptiveness of the class names. However, these improvements are restricted to the text-based classifier only, and the query visual features are not considered. In this paper, we propose a framework which combines pre-trained discriminative VLMs with pre-trained generative video-to-text and text-to-text models. We introduce two key modifications to the standard zero-shot setting. First, we propose language-guided visual feature enhancement and employ a video-to-text model to convert the query video to its descriptive form. The resulting descriptions contain vital visual cues of the query video, such as what objects are present and their spatio-temporal interactions. These descriptive cues provide additional semantic knowledge to VLMs to enhance their zeroshot performance. Second, we propose video-specific prompts to LLMs to generate more meaningful descriptions to enrich class label representations. Specifically, we introduce prompt techniques to create a Tree Hierarchy of Categories for class names, offering a higher-level action context for additional visual cues, We demonstrate the effectiveness of our approach in video understanding across three different zero-shot settings: 1) video action recognition, 2) video-to-text and textto-video retrieval, and 3) time-sensitive video tasks. Consistent improvements across multiple benchmarks and with various VLMs demonstrate the effectiveness of our proposed framework. Our code will be made publicly available.

Aisha Urooj Khan, Hilde Kuehne, Chuang Gan et al.

Transformers for visual-language representation learning have been getting a lot of interest and shown tremendous performance on visual question answering (VQA) and grounding. But most systems that show good performance of those tasks still rely on pre-trained object detectors during training, which limits their applicability to the object classes available for those detectors. To mitigate this limitation, the following paper focuses on the problem of weakly supervised grounding in context of visual question answering in transformers. The approach leverages capsules by grouping each visual token in the visual encoder and uses activations from language self-attention layers as a text-guided selection module to mask those capsules before they are forwarded to the next layer. We evaluate our approach on the challenging GQA as well as VQA-HAT dataset for VQA grounding. Our experiments show that: while removing the information of masked objects from standard transformer architectures leads to a significant drop in performance, the integration of capsules significantly improves the grounding ability of such systems and provides new state-of-the-art results compared to other approaches in the field.

Ishan Rajendrakumar Dave, Fabian Caba Heilbron, Mubarak Shah et al.

Temporal video alignment aims to synchronize the key events like object interactions or action phase transitions in two videos. Such methods could benefit various video editing, processing, and understanding tasks. However, existing approaches operate under the restrictive assumption that a suitable video pair for alignment is given, significantly limiting their broader applicability. To address this, we re-pose temporal alignment as a search problem and introduce the task of Alignable Video Retrieval (AVR). Given a query video, our approach can identify well-alignable videos from a large collection of clips and temporally synchronize them to the query. To achieve this, we make three key contributions: 1) we introduce DRAQ, a video alignability indicator to identify and re-rank the best alignable video from a set of candidates; 2) we propose an effective and generalizable frame-level video feature design to improve the alignment performance of several off-the-shelf feature representations, and 3) we propose a novel benchmark and evaluation protocol for AVR using cycle-consistency metrics. Our experiments on 3 datasets, including large-scale Kinetics700, demonstrate the effectiveness of our approach in identifying alignable video pairs from diverse datasets. Project Page: https://daveishan.github.io/avr-webpage/.

Soumyabrata Dey, Ravishankar Rao, Mubarak Shah

Attention Deficit Hyperactive Disorder (ADHD) is a common behavioral problem affecting children. In this work, we investigate the automatic classification of ADHD subjects using the resting state Functional Magnetic Resonance Imaging (fMRI) sequences of the brain. We show that the brain can be modeled as a functional network, and certain properties of the networks differ in ADHD subjects from control subjects. We compute the pairwise correlation of brain voxels' activity over the time frame of the experimental protocol which helps to model the function of a brain as a network. Different network features are computed for each of the voxels constructing the network. The concatenation of the network features of all the voxels in a brain serves as the feature vector. Feature vectors from a set of subjects are then used to train a PCA-LDA (principal component analysis-linear discriminant analysis) based classifier. We hypothesized that ADHD-related differences lie in some specific regions of the brain and using features only from those regions is sufficient to discriminate ADHD and control subjects. We propose a method to create a brain mask that includes the useful regions only and demonstrate that using the feature from the masked regions improves classification accuracy on the test data set. We train our classifier with 776 subjects and test on 171 subjects provided by The Neuro Bureau for the ADHD-200 challenge. We demonstrate the utility of graph-motif features, specifically the maps that represent the frequency of participation of voxels in network cycles of length 3. The best classification performance (69.59%) is achieved using 3-cycle map features with masking. Our proposed approach holds promise in being able to diagnose and understand the disorder.

James Beetham, Navid Kardan, Ajmal Mian et al.

Existing data-free model stealing methods use a generator to produce samples in order to train a student model to match the target model outputs. To this end, the two main challenges are estimating gradients of the target model without access to its parameters, and generating a diverse set of training samples that thoroughly explores the input space. We propose a Dual Student method where two students are symmetrically trained in order to provide the generator a criterion to generate samples that the two students disagree on. On one hand, disagreement on a sample implies at least one student has classified the sample incorrectly when compared to the target model. This incentive towards disagreement implicitly encourages the generator to explore more diverse regions of the input space. On the other hand, our method utilizes gradients of student models to indirectly estimate gradients of the target model. We show that this novel training objective for the generator network is equivalent to optimizing a lower bound on the generator's loss if we had access to the target model gradients. We show that our new optimization framework provides more accurate gradient estimation of the target model and better accuracies on benchmark classification datasets. Additionally, our approach balances improved query efficiency with training computation cost. Finally, we demonstrate that our method serves as a better proxy model for transfer-based adversarial attacks than existing data-free model stealing methods.

Yuning Cui, Syed Waqas Zamir, Salman Khan et al.

In the image acquisition process, various forms of degradation, including noise, haze, and rain, are frequently introduced. These degradations typically arise from the inherent limitations of cameras or unfavorable ambient conditions. To recover clean images from degraded versions, numerous specialized restoration methods have been developed, each targeting a specific type of degradation. Recently, all-in-one algorithms have garnered significant attention by addressing different types of degradations within a single model without requiring prior information of the input degradation type. However, these methods purely operate in the spatial domain and do not delve into the distinct frequency variations inherent to different degradation types. To address this gap, we propose an adaptive all-in-one image restoration network based on frequency mining and modulation. Our approach is motivated by the observation that different degradation types impact the image content on different frequency subbands, thereby requiring different treatments for each restoration task. Specifically, we first mine low- and high-frequency information from the input features, guided by the adaptively decoupled spectra of the degraded image. The extracted features are then modulated by a bidirectional operator to facilitate interactions between different frequency components. Finally, the modulated features are merged into the original input for a progressively guided restoration. With this approach, the model achieves adaptive reconstruction by accentuating the informative frequency subbands according to different input degradations. Extensive experiments demonstrate that the proposed method achieves state-of-the-art performance on different image restoration tasks, including denoising, dehazing, deraining, motion deblurring, and low-light image enhancement. Our code is available at https://github.com/c-yn/AdaIR.

Daochang Liu, Qiyue Li, AnhDung Dinh et al.

Temporal action segmentation is crucial for understanding long-form videos. Previous works on this task commonly adopt an iterative refinement paradigm by using multi-stage models. We propose a novel framework via denoising diffusion models, which nonetheless shares the same inherent spirit of such iterative refinement. In this framework, action predictions are iteratively generated from random noise with input video features as conditions. To enhance the modeling of three striking characteristics of human actions, including the position prior, the boundary ambiguity, and the relational dependency, we devise a unified masking strategy for the conditioning inputs in our framework. Extensive experiments on three benchmark datasets, i.e., GTEA, 50Salads, and Breakfast, are performed and the proposed method achieves superior or comparable results to state-of-the-art methods, showing the effectiveness of a generative approach for action segmentation.

Jyoti Kini, Sarah Fleischer, Ishan Dave et al.

In this work, we propose an ensemble modeling approach for multimodal action recognition. We independently train individual modality models using a variant of focal loss tailored to handle the long-tailed distribution of the MECCANO [21] dataset. Based on the underlying principle of focal loss, which captures the relationship between tail (scarce) classes and their prediction difficulties, we propose an exponentially decaying variant of focal loss for our current task. It initially emphasizes learning from the hard misclassified examples and gradually adapts to the entire range of examples in the dataset. This annealing process encourages the model to strike a balance between focusing on the sparse set of hard samples, while still leveraging the information provided by the easier ones. Additionally, we opt for the late fusion strategy to combine the resultant probability distributions from RGB and Depth modalities for final action prediction. Experimental evaluations on the MECCANO dataset demonstrate the effectiveness of our approach.

Ishan Rajendrakumar Dave, Mamshad Nayeem Rizve, Mubarak Shah

Real-life applications of action recognition often require a fine-grained understanding of subtle movements, e.g., in sports analytics, user interactions in AR/VR, and surgical videos. Although fine-grained actions are more costly to annotate, existing semi-supervised action recognition has mainly focused on coarse-grained action recognition. Since fine-grained actions are more challenging due to the absence of scene bias, classifying these actions requires an understanding of action-phases. Hence, existing coarse-grained semi-supervised methods do not work effectively. In this work, we for the first time thoroughly investigate semi-supervised fine-grained action recognition (FGAR). We observe that alignment distances like dynamic time warping (DTW) provide a suitable action-phase-aware measure for comparing fine-grained actions, a concept previously unexploited in FGAR. However, since regular DTW distance is pairwise and assumes strict alignment between pairs, it is not directly suitable for classifying fine-grained actions. To utilize such alignment distances in a limited-label setting, we propose an Alignability-Verification-based Metric learning technique to effectively discriminate between fine-grained action pairs. Our learnable alignability score provides a better phase-aware measure, which we use to refine the pseudo-labels of the primary video encoder. Our collaborative pseudo-labeling-based framework `\textit{FinePseudo}' significantly outperforms prior methods on four fine-grained action recognition datasets: Diving48, FineGym99, FineGym288, and FineDiving, and shows improvement on existing coarse-grained datasets: Kinetics400 and Something-SomethingV2. We also demonstrate the robustness of our collaborative pseudo-labeling in handling novel unlabeled classes in open-world semi-supervised setups. Project Page: https://daveishan.github.io/finepsuedo-webpage/.

Jyoti Kini, Sarah Fleischer, Ishan Dave et al.

Action recognition from an egocentric viewpoint is a crucial perception task in robotics and enables a wide range of human-robot interactions. While most computer vision approaches prioritize the RGB camera, the Depth modality - which can further amplify the subtleties of actions from an egocentric perspective - remains underexplored. Our work focuses on recognizing actions from egocentric RGB and Depth modalities in an industry-like environment. To study this problem, we consider the recent MECCANO dataset, which provides a wide range of assembling actions. Our framework is based on the 3D Video SWIN Transformer to encode both RGB and Depth modalities effectively. To address the inherent skewness in real-world multimodal action occurrences, we propose a training strategy using an exponentially decaying variant of the focal loss modulating factor. Additionally, to leverage the information in both RGB and Depth modalities, we opt for late fusion to combine the predictions from each modality. We thoroughly evaluate our method on the action recognition task of the MECCANO dataset, and it significantly outperforms the prior work. Notably, our method also secured first place at the multimodal action recognition challenge at ICIAP 2023.