Nuno Vasconcelos

Tz-Ying Wu, Gurumurthy Swaminathan, Zhizhong Li et al. · amazon-science

Class-incremental learning (CIL) has been widely studied under the setting of starting from a small number of classes (base classes). Instead, we explore an understudied real-world setting of CIL that starts with a strong model pre-trained on a large number of base classes. We hypothesize that a strong base model can provide a good representation for novel classes and incremental learning can be done with small adaptations. We propose a 2-stage training scheme, i) feature augmentation -- cloning part of the backbone and fine-tuning it on the novel data, and ii) fusion -- combining the base and novel classifiers into a unified classifier. Experiments show that the proposed method significantly outperforms state-of-the-art CIL methods on the large-scale ImageNet dataset (e.g. +10% overall accuracy than the best). We also propose and analyze understudied practical CIL scenarios, such as base-novel overlap with distribution shift. Our proposed method is robust and generalizes to all analyzed CIL settings. Code is available at https://github.com/amazon-research/sp-cil.

Aditya Sarkar, Yi Li, Zihao Wang et al.

A new framework for the estimation of the complexity posed by video-question pairs to video-LLMs, Video Attribute-Based Complexity (VideoABC), is proposed. Video complexity is defined as the probability of failure of a video-LLM for a given video-question pair. VideoABC is a non-parametric complexity measure, using a reference video dataset and a pre-defined vocabulary of video attributes informative of complexity, \eg the scene complexity or the speed of the video event informative of the question. In a training phase, reference videos are projected into the space of these attributes, which is then quantized. The expected ABC of each quantization cell is then computed. Given a new video and its projection into the attribute space, complexity is estimated by the expected ABC of the associated quantization cell. To enable the use of VideoABC with small reference video datasets, two quantizers are combined: a k-means quantizer that enables accurate complexity estimates for samples in the distribution of the reference dataset and a universal lattice quantizer that guarantees generalization to out-of-distribution samples. A synthetic video generation procedure, inspired by target-distractor manipulations of psychophysics studies, is proposed to populate the cells of the lattice quantizer during training, enabling the computation of their expected ABCs. Experimental results show that VideoABCis effective even with very low-dimensional attribute representations, substantially outperforming approaches like `video-LLM as judge' with much less complexity. Finally, the explainable nature of the VideoABC score, in terms of well-defined attributes, is shown to provide insights on how the attribute composition of benchmarks affects their complexity.

Chih-Hui Ho, Srikar Appalaraju, Bhavan Jasani et al. · amazon-science

We present YORO - a multi-modal transformer encoder-only architecture for the Visual Grounding (VG) task. This task involves localizing, in an image, an object referred via natural language. Unlike the recent trend in the literature of using multi-stage approaches that sacrifice speed for accuracy, YORO seeks a better trade-off between speed an accuracy by embracing a single-stage design, without CNN backbone. YORO consumes natural language queries, image patches, and learnable detection tokens and predicts coordinates of the referred object, using a single transformer encoder. To assist the alignment between text and visual objects, a novel patch-text alignment loss is proposed. Extensive experiments are conducted on 5 different datasets with ablations on architecture design choices. YORO is shown to support real-time inference and outperform all approaches in this class (single-stage methods) by large margins. It is also the fastest VG model and achieves the best speed/accuracy trade-off in the literature.

Zhiyuan Hu, Yunsheng Li, Jiancheng Lyu et al.

The problem of class incremental learning (CIL) is considered. State-of-the-art approaches use a dynamic architecture based on network expansion (NE), in which a task expert is added per task. While effective from a computational standpoint, these methods lead to models that grow quickly with the number of tasks. A new NE method, dense network expansion (DNE), is proposed to achieve a better trade-off between accuracy and model complexity. This is accomplished by the introduction of dense connections between the intermediate layers of the task expert networks, that enable the transfer of knowledge from old to new tasks via feature sharing and reusing. This sharing is implemented with a cross-task attention mechanism, based on a new task attention block (TAB), that fuses information across tasks. Unlike traditional attention mechanisms, TAB operates at the level of the feature mixing and is decoupled with spatial attentions. This is shown more effective than a joint spatial-and-task attention for CIL. The proposed DNE approach can strictly maintain the feature space of old classes while growing the network and feature scale at a much slower rate than previous methods. In result, it outperforms the previous SOTA methods by a margin of 4\% in terms of accuracy, with similar or even smaller model scale.

Chih-Hui Ho, Nuno Vasconcelos

The problem of adversarial defenses for image classification, where the goal is to robustify a classifier against adversarial examples, is considered. Inspired by the hypothesis that these examples lie beyond the natural image manifold, a novel aDversarIal defenSe with local impliCit functiOns (DISCO) is proposed to remove adversarial perturbations by localized manifold projections. DISCO consumes an adversarial image and a query pixel location and outputs a clean RGB value at the location. It is implemented with an encoder and a local implicit module, where the former produces per-pixel deep features and the latter uses the features in the neighborhood of query pixel for predicting the clean RGB value. Extensive experiments demonstrate that both DISCO and its cascade version outperform prior defenses, regardless of whether the defense is known to the attacker. DISCO is also shown to be data and parameter efficient and to mount defenses that transfers across datasets, classifiers and attacks.

Yi Li, Rameswar Panda, Yoon Kim et al.

Designing better machine translation systems by considering auxiliary inputs such as images has attracted much attention in recent years. While existing methods show promising performance over the conventional text-only translation systems, they typically require paired text and image as input during inference, which limits their applicability to real-world scenarios. In this paper, we introduce a visual hallucination framework, called VALHALLA, which requires only source sentences at inference time and instead uses hallucinated visual representations for multimodal machine translation. In particular, given a source sentence an autoregressive hallucination transformer is used to predict a discrete visual representation from the input text, and the combined text and hallucinated representations are utilized to obtain the target translation. We train the hallucination transformer jointly with the translation transformer using standard backpropagation with cross-entropy losses while being guided by an additional loss that encourages consistency between predictions using either ground-truth or hallucinated visual representations. Extensive experiments on three standard translation datasets with a diverse set of language pairs demonstrate the effectiveness of our approach over both text-only baselines and state-of-the-art methods. Project page: http://www.svcl.ucsd.edu/projects/valhalla.

Jiteng Mu, Shen Sang, Nuno Vasconcelos et al.

While NeRF-based human representations have shown impressive novel view synthesis results, most methods still rely on a large number of images / views for training. In this work, we propose a novel animatable NeRF called ActorsNeRF. It is first pre-trained on diverse human subjects, and then adapted with few-shot monocular video frames for a new actor with unseen poses. Building on previous generalizable NeRFs with parameter sharing using a ConvNet encoder, ActorsNeRF further adopts two human priors to capture the large human appearance, shape, and pose variations. Specifically, in the encoded feature space, we will first align different human subjects in a category-level canonical space, and then align the same human from different frames in an instance-level canonical space for rendering. We quantitatively and qualitatively demonstrate that ActorsNeRF significantly outperforms the existing state-of-the-art on few-shot generalization to new people and poses on multiple datasets. Project Page: https://jitengmu.github.io/ActorsNeRF/

Yi Li, Kyle Min, Subarna Tripathi et al.

Do video-text transformers learn to model temporal relationships across frames? Despite their immense capacity and the abundance of multimodal training data, recent work has revealed the strong tendency of video-text models towards frame-based spatial representations, while temporal reasoning remains largely unsolved. In this work, we identify several key challenges in temporal learning of video-text transformers: the spatiotemporal trade-off from limited network size; the curse of dimensionality for multi-frame modeling; and the diminishing returns of semantic information by extending clip length. Guided by these findings, we propose SViTT, a sparse video-text architecture that performs multi-frame reasoning with significantly lower cost than naive transformers with dense attention. Analogous to graph-based networks, SViTT employs two forms of sparsity: edge sparsity that limits the query-key communications between tokens in self-attention, and node sparsity that discards uninformative visual tokens. Trained with a curriculum which increases model sparsity with the clip length, SViTT outperforms dense transformer baselines on multiple video-text retrieval and question answering benchmarks, with a fraction of computational cost. Project page: http://svcl.ucsd.edu/projects/svitt.

Tz-Ying Wu, Chih-Hui Ho, Nuno Vasconcelos

Visual-language foundation models, like CLIP, learn generalized representations that enable zero-shot open-set classification. Few-shot adaptation methods, based on prompt tuning, have been shown to further improve performance on downstream datasets. However, these methods do not fare well in the taxonomic open set (TOS) setting, where the classifier is asked to make predictions from label sets across different levels of semantic granularity. Frequently, they infer incorrect labels at coarser taxonomic class levels, even when the inference at the leaf level (original class labels) is correct. To address this problem, we propose a prompt tuning technique that calibrates the hierarchical consistency of model predictions. A set of metrics of hierarchical consistency, the Hierarchical Consistent Accuracy (HCA) and the Mean Treecut Accuracy (MTA), are first proposed to evaluate TOS model performance. A new Prompt Tuning for Hierarchical Consistency (ProTeCt) technique is then proposed to calibrate classification across label set granularities. Results show that ProTeCt can be combined with existing prompt tuning methods to significantly improve TOS classification without degrading the leaf level classification performance.

Zhiyuan Hu, Jiancheng Lyu, Dashan Gao et al.

Continual learning (CL) has attracted increasing attention in the recent past. It aims to mimic the human ability to learn new concepts without catastrophic forgetting. While existing CL methods accomplish this to some extent, they are still prone to semantic drift of the learned feature space. Foundation models, which are endowed with a robust feature representation, learned from very large datasets, provide an interesting substrate for the solution of the CL problem. Recent work has also shown that they can be adapted to specific tasks by prompt tuning techniques that leave the generality of the representation mostly unscathed. An open question is, however, how to learn both prompts that are task specific and prompts that are global, i.e. capture cross-task information. In this work, we propose the Prompt Of Prompts (POP) model, which addresses this goal by progressively learning a group of task-specified prompts and a group of global prompts, denoted as POP, to integrate information from the former. We show that a foundation model equipped with POP learning is able to outperform classic CL methods by a significant margin. Moreover, as prompt tuning only requires a small set of training samples, POP is able to perform CL in the few-shot setting, while still outperforming competing methods trained on the entire dataset.

Alakh Desai, Tz-Ying Wu, Subarna Tripathi et al.

Research in scene graph generation (SGG) usually considers two-stage models, that is, detecting a set of entities, followed by combining them and labeling all possible relationships. While showing promising results, the pipeline structure induces large parameter and computation overhead, and typically hinders end-to-end optimizations. To address this, recent research attempts to train single-stage models that are computationally efficient. With the advent of DETR, a set based detection model, one-stage models attempt to predict a set of subject-predicate-object triplets directly in a single shot. However, SGG is inherently a multi-task learning problem that requires modeling entity and predicate distributions simultaneously. In this paper, we propose Transformers with conditional queries for SGG, namely, TraCQ with a new formulation for SGG that avoids the multi-task learning problem and the combinatorial entity pair distribution. We employ a DETR-based encoder-decoder design and leverage conditional queries to significantly reduce the entity label space as well, which leads to 20% fewer parameters compared to state-of-the-art single-stage models. Experimental results show that TraCQ not only outperforms existing single-stage scene graph generation methods, it also beats many state-of-the-art two-stage methods on the Visual Genome dataset, yet is capable of end-to-end training and faster inference.

Mark Tenzer, Zeeshan Rasheed, Khurram Shafique et al.

A need to understand and predict vehicles' behavior underlies both public and private goals in the transportation domain, including urban planning and management, ride-sharing services, and intelligent transportation systems. Individuals' preferences and intended destinations vary throughout the day, week, and year: for example, bars are most popular in the evenings, and beaches are most popular in the summer. Despite this principle, we note that recent studies on a popular benchmark dataset from Porto, Portugal have found, at best, only marginal improvements in predictive performance from incorporating temporal information. We propose an approach based on hypernetworks, a variant of meta-learning ("learning to learn") in which a neural network learns to change its own weights in response to an input. In our case, the weights responsible for destination prediction vary with the metadata, in particular the time, of the input trajectory. The time-conditioned weights notably improve the model's error relative to ablation studies and comparable prior work, and we confirm our hypothesis that knowledge of time should improve prediction of a vehicle's intended destination.

Yunsheng Li, Yinpeng Chen, Xiyang Dai et al.

The complexity-precision trade-off of an object detector is a critical problem for resource constrained vision tasks. Previous works have emphasized detectors implemented with efficient backbones. The impact on this trade-off of proposal processing by the detection head is investigated in this work. It is hypothesized that improved detection efficiency requires a paradigm shift, towards the unequal processing of proposals, assigning more computation to good proposals than poor ones. This results in better utilization of available computational budget, enabling higher accuracy for the same FLOPS. We formulate this as a learning problem where the goal is to assign operators to proposals, in the detection head, so that the total computational cost is constrained and the precision is maximized. The key finding is that such matching can be learned as a function that maps each proposal embedding into a one-hot code over operators. While this function induces a complex dynamic network routing mechanism, it can be implemented by a simple MLP and learned end-to-end with off-the-shelf object detectors. This 'dynamic proposal processing' (DPP) is shown to outperform state-of-the-art end-to-end object detectors (DETR, Sparse R-CNN) by a clear margin for a given computational complexity.

Yuzhao Chen, Zonghuan Li, Zhiyuan Hu et al.

The problem of continual learning has attracted rising attention in recent years. However, few works have questioned the commonly used learning setup, based on a task curriculum of random class. This differs significantly from human continual learning, which is guided by taxonomic curricula. In this work, we propose the Taxonomic Class Incremental Learning (TCIL) problem. In TCIL, the task sequence is organized based on a taxonomic class tree. We unify existing approaches to CIL and taxonomic learning as parameter inheritance schemes and introduce a new such scheme for the TCIL learning. This enables the incremental transfer of knowledge from ancestor to descendant class of a class taxonomy through parameter inheritance. Experiments on CIFAR-100 and ImageNet-100 show the effectiveness of the proposed TCIL method, which outperforms existing SOTA methods by 2% in terms of final accuracy on CIFAR-100 and 3% on ImageNet-100.

Yuwei Zhang, Chih-Hui Ho, Nuno Vasconcelos

The problem of realistic VQA (RVQA), where a model has to reject unanswerable questions (UQs) and answer answerable ones (AQs), is studied. We first point out 2 drawbacks in current RVQA research, where (1) datasets contain too many unchallenging UQs and (2) a large number of annotated UQs are required for training. To resolve the first drawback, we propose a new testing dataset, RGQA, which combines AQs from an existing VQA dataset with around 29K human-annotated UQs. These UQs consist of both fine-grained and coarse-grained image-question pairs generated with 2 approaches: CLIP-based and Perturbation-based. To address the second drawback, we introduce an unsupervised training approach. This combines pseudo UQs obtained by randomly pairing images and questions, with an RoI Mixup procedure to generate more fine-grained pseudo UQs, and model ensembling to regularize model confidence. Experiments show that using pseudo UQs significantly outperforms RVQA baselines. RoI Mixup and model ensembling further increase the gain. Finally, human evaluation reveals a performance gap between humans and models, showing that more RVQA research is needed.

Sepehr Dehdashtian, Ruozhen He, Yi Li et al.

Computer vision systems have witnessed rapid progress over the past two decades due to multiple advances in the field. As these systems are increasingly being deployed in high-stakes real-world applications, there is a dire need to ensure that they do not propagate or amplify any discriminatory tendencies in historical or human-curated data or inadvertently learn biases from spurious correlations. This paper presents a comprehensive survey on fairness that summarizes and sheds light on ongoing trends and successes in the context of computer vision. The topics we discuss include 1) The origin and technical definitions of fairness drawn from the wider fair machine learning literature and adjacent disciplines. 2) Work that sought to discover and analyze biases in computer vision systems. 3) A summary of methods proposed to mitigate bias in computer vision systems in recent years. 4) A comprehensive summary of resources and datasets produced by researchers to measure, analyze, and mitigate bias and enhance fairness. 5) Discussion of the field's success, continuing trends in the context of multimodal foundation and generative models, and gaps that still need to be addressed. The presented characterization should help researchers understand the importance of identifying and mitigating bias in computer vision and the state of the field and identify potential directions for future research.

Aditya Sarkar, Yi Li, Jiacheng Cheng et al.

Selective prediction aims to endow predictors with a reject option, to avoid low confidence predictions. However, existing literature has primarily focused on closed-set tasks, such as visual question answering with predefined options or fixed-category classification. This paper considers selective prediction for visual language foundation models, addressing a taxonomy of tasks ranging from closed to open set and from finite to unbounded vocabularies, as in image captioning. We seek training-free approaches of low-complexity, applicable to any foundation model and consider methods based on external vision-language model embeddings, like CLIP. This is denoted as Plug-and-Play Selective Prediction (PaPSP). We identify two key challenges: (1) instability of the visual-language representations, leading to high variance in image-text embeddings, and (2) poor calibration of similarity scores. To address these issues, we propose a memory augmented PaPSP (MA-PaPSP) model, which augments PaPSP with a retrieval dataset of image-text pairs. This is leveraged to reduce embedding variance by averaging retrieved nearest-neighbor pairs and is complemented by the use of contrastive normalization to improve score calibration. Through extensive experiments on multiple datasets, we show that MA-PaPSP outperforms PaPSP and other selective prediction baselines for selective captioning, image-text matching, and fine-grained classification. Code is publicly available at https://github.com/kingston-aditya/MA-PaPSP.

Maijunxian Wang, Ruisi Wang, Juyi Lin et al.

Rapid progress in video models has largely focused on visual quality, leaving their reasoning capabilities underexplored. Video reasoning grounds intelligence in spatiotemporally consistent visual environments that go beyond what text can naturally capture, enabling intuitive reasoning over spatiotemporal structure such as continuity, interaction, and causality. However, systematically studying video reasoning and its scaling behavior is hindered by the lack of large-scale training data. To address this gap, we introduce the Very Big Video Reasoning (VBVR) Dataset, an unprecedentedly large-scale resource spanning 200 curated reasoning tasks following a principled taxonomy and over one million video clips, approximately three orders of magnitude larger than existing datasets. We further present VBVR-Bench, a verifiable evaluation framework that moves beyond model-based judging by incorporating rule-based, human-aligned scorers, enabling reproducible and interpretable diagnosis of video reasoning capabilities. Leveraging the VBVR suite, we conduct one of the first large-scale scaling studies of video reasoning and observe early signs of emergent generalization to unseen reasoning tasks. Together, VBVR lays a foundation for the next stage of research in generalizable video reasoning. The data, benchmark toolkit, and models are publicly available at https://video-reason.com/ .

Yijiang Li, Genpei Zhang, Jiacheng Cheng et al.

While the rapid proliferation of wearable cameras has raised significant concerns about egocentric video privacy, prior work has largely overlooked the unique privacy threats posed to the camera wearer. This work investigates the core question: How much privacy information about the camera wearer can be inferred from their first-person view videos? We introduce EgoPrivacy, the first large-scale benchmark for the comprehensive evaluation of privacy risks in egocentric vision. EgoPrivacy covers three types of privacy (demographic, individual, and situational), defining seven tasks that aim to recover private information ranging from fine-grained (e.g., wearer's identity) to coarse-grained (e.g., age group). To further emphasize the privacy threats inherent to egocentric vision, we propose Retrieval-Augmented Attack, a novel attack strategy that leverages ego-to-exo retrieval from an external pool of exocentric videos to boost the effectiveness of demographic privacy attacks. An extensive comparison of the different attacks possible under all threat models is presented, showing that private information of the wearer is highly susceptible to leakage. For instance, our findings indicate that foundation models can effectively compromise wearer privacy even in zero-shot settings by recovering attributes such as identity, scene, gender, and race with 70-80% accuracy. Our code and data are available at https://github.com/williamium3000/ego-privacy.

Tz-Ying Wu, Kyle Min, Subarna Tripathi et al.

Video understanding typically requires fine-tuning the large backbone when adapting to new domains. In this paper, we leverage the egocentric video foundation models (Ego-VFMs) based on video-language pre-training and propose a parameter-efficient adaptation for egocentric video tasks, namely Ego-VPA. It employs a local sparse approximation for each video frame/text feature using the basis prompts, and the selected basis prompts are used to synthesize video/text prompts. Since the basis prompts are shared across frames and modalities, it models context fusion and cross-modal transfer in an efficient fashion. Experiments show that Ego-VPA excels in lightweight adaptation (with only 0.84% learnable parameters), largely improving over baselines and reaching the performance of full fine-tuning.

Deepak Sridhar, Nuno Vasconcelos

Diffusion models have recently surpassed GANs in image synthesis and editing, offering superior image quality and diversity. However, achieving precise control over attributes in generated images remains a challenge. Concept Sliders introduced a method for fine-grained image control and editing by learning concepts (attributes/objects). However, this approach adds parameters and increases inference time due to the loading and unloading of Low-Rank Adapters (LoRAs) used for learning concepts. These adapters are model-specific and require retraining for different architectures, such as Stable Diffusion (SD) v1.5 and SD-XL. In this paper, we propose a straightforward textual inversion method to learn concepts through text embeddings, which are generalizable across models that share the same text encoder, including different versions of the SD model. We refer to our method as Prompt Sliders. Besides learning new concepts, we also show that Prompt Sliders can be used to erase undesirable concepts such as artistic styles or mature content. Our method is 30% faster than using LoRAs because it eliminates the need to load and unload adapters and introduces no additional parameters aside from the target concept text embedding. Each concept embedding only requires 3KB of storage compared to the 8922KB or more required for each LoRA adapter, making our approach more computationally efficient. Project Page: https://deepaksridhar.github.io/promptsliders.github.io/

Pei Wang, Zhaowei Cai, Hao Yang et al.

We consider the problem of omni-supervised object detection, which can use unlabeled, fully labeled and weakly labeled annotations, such as image tags, counts, points, etc., for object detection. This is enabled by a unified architecture, Omni-DETR, based on the recent progress on student-teacher framework and end-to-end transformer based object detection. Under this unified architecture, different types of weak labels can be leveraged to generate accurate pseudo labels, by a bipartite matching based filtering mechanism, for the model to learn. In the experiments, Omni-DETR has achieved state-of-the-art results on multiple datasets and settings. And we have found that weak annotations can help to improve detection performance and a mixture of them can achieve a better trade-off between annotation cost and accuracy than the standard complete annotation. These findings could encourage larger object detection datasets with mixture annotations. The code is available at https://github.com/amazon-research/omni-detr.

Yunsheng Li, Yinpeng Chen, Xiyang Dai et al.

This paper aims at addressing the problem of substantial performance degradation at extremely low computational cost (e.g. 5M FLOPs on ImageNet classification). We found that two factors, sparse connectivity and dynamic activation function, are effective to improve the accuracy. The former avoids the significant reduction of network width, while the latter mitigates the detriment of reduction in network depth. Technically, we propose micro-factorized convolution, which factorizes a convolution matrix into low rank matrices, to integrate sparse connectivity into convolution. We also present a new dynamic activation function, named Dynamic Shift Max, to improve the non-linearity via maxing out multiple dynamic fusions between an input feature map and its circular channel shift. Building upon these two new operators, we arrive at a family of networks, named MicroNet, that achieves significant performance gains over the state of the art in the low FLOP regime. For instance, under the constraint of 12M FLOPs, MicroNet achieves 59.4\% top-1 accuracy on ImageNet classification, outperforming MobileNetV3 by 9.6\%. Source code is at \href{https://github.com/liyunsheng13/micronet}{https://github.com/liyunsheng13/micronet}.

Yunsheng Li, Lu Yuan, Yinpeng Chen et al.

Recent works of multi-source domain adaptation focus on learning a domain-agnostic model, of which the parameters are static. However, such a static model is difficult to handle conflicts across multiple domains, and suffers from a performance degradation in both source domains and target domain. In this paper, we present dynamic transfer to address domain conflicts, where the model parameters are adapted to samples. The key insight is that adapting model across domains is achieved via adapting model across samples. Thus, it breaks down source domain barriers and turns multi-source domains into a single-source domain. This also simplifies the alignment between source and target domains, as it only requires the target domain to be aligned with any part of the union of source domains. Furthermore, we find dynamic transfer can be simply modeled by aggregating residual matrices and a static convolution matrix. Experimental results show that, without using domain labels, our dynamic transfer outperforms the state-of-the-art method by more than 3% on the large multi-source domain adaptation datasets -- DomainNet. Source code is at https://github.com/liyunsheng13/DRT.

Yunsheng Li, Yinpeng Chen, Xiyang Dai et al.

Recent research in dynamic convolution shows substantial performance boost for efficient CNNs, due to the adaptive aggregation of K static convolution kernels. It has two limitations: (a) it increases the number of convolutional weights by K-times, and (b) the joint optimization of dynamic attention and static convolution kernels is challenging. In this paper, we revisit it from a new perspective of matrix decomposition and reveal the key issue is that dynamic convolution applies dynamic attention over channel groups after projecting into a higher dimensional latent space. To address this issue, we propose dynamic channel fusion to replace dynamic attention over channel groups. Dynamic channel fusion not only enables significant dimension reduction of the latent space, but also mitigates the joint optimization difficulty. As a result, our method is easier to train and requires significantly fewer parameters without sacrificing accuracy. Source code is at https://github.com/liyunsheng13/dcd.

Chih-Hui Ho, Bo Liu, Tz-Ying Wu et al.

Multiview recognition has been well studied in the literature and achieves decent performance in object recognition and retrieval task. However, most previous works rely on supervised learning and some impractical underlying assumptions, such as the availability of all views in training and inference time. In this work, the problem of multiview self-supervised learning (MV-SSL) is investigated, where only image to object association is given. Given this setup, a novel surrogate task for self-supervised learning is proposed by pursuing "object invariant" representation. This is solved by randomly selecting an image feature of an object as object prototype, accompanied with multiview consistency regularization, which results in view invariant stochastic prototype embedding (VISPE). Experiments shows that the recognition and retrieval results using VISPE outperform that of other self-supervised learning methods on seen and unseen data. VISPE can also be applied to semi-supervised scenario and demonstrates robust performance with limited data available. Code is available at https://github.com/chihhuiho/VISPE

Zhaowei Cai, Nuno Vasconcelos

In object detection, the intersection over union (IoU) threshold is frequently used to define positives/negatives. The threshold used to train a detector defines its \textit{quality}. While the commonly used threshold of 0.5 leads to noisy (low-quality) detections, detection performance frequently degrades for larger thresholds. This paradox of high-quality detection has two causes: 1) overfitting, due to vanishing positive samples for large thresholds, and 2) inference-time quality mismatch between detector and test hypotheses. A multi-stage object detection architecture, the Cascade R-CNN, composed of a sequence of detectors trained with increasing IoU thresholds, is proposed to address these problems. The detectors are trained sequentially, using the output of a detector as training set for the next. This resampling progressively improves hypotheses quality, guaranteeing a positive training set of equivalent size for all detectors and minimizing overfitting. The same cascade is applied at inference, to eliminate quality mismatches between hypotheses and detectors. An implementation of the Cascade R-CNN without bells or whistles achieves state-of-the-art performance on the COCO dataset, and significantly improves high-quality detection on generic and specific object detection datasets, including VOC, KITTI, CityPerson, and WiderFace. Finally, the Cascade R-CNN is generalized to instance segmentation, with nontrivial improvements over the Mask R-CNN. To facilitate future research, two implementations are made available at \url{https://github.com/zhaoweicai/cascade-rcnn} (Caffe) and \url{https://github.com/zhaoweicai/Detectron-Cascade-RCNN} (Detectron).

Yunsheng Li, Lu Yuan, Nuno Vasconcelos

Domain adaptation for semantic image segmentation is very necessary since manually labeling large datasets with pixel-level labels is expensive and time consuming. Existing domain adaptation techniques either work on limited datasets, or yield not so good performance compared with supervised learning. In this paper, we propose a novel bidirectional learning framework for domain adaptation of segmentation. Using the bidirectional learning, the image translation model and the segmentation adaptation model can be learned alternatively and promote to each other. Furthermore, we propose a self-supervised learning algorithm to learn a better segmentation adaptation model and in return improve the image translation model. Experiments show that our method is superior to the state-of-the-art methods in domain adaptation of segmentation with a big margin. The source code is available at https://github.com/liyunsheng13/BDL.

Yi Li, Nuno Vasconcelos

Modern machine learning datasets can have biases for certain representations that are leveraged by algorithms to achieve high performance without learning to solve the underlying task. This problem is referred to as "representation bias". The question of how to reduce the representation biases of a dataset is investigated and a new dataset REPresentAtion bIas Removal (REPAIR) procedure is proposed. This formulates bias minimization as an optimization problem, seeking a weight distribution that penalizes examples easy for a classifier built on a given feature representation. Bias reduction is then equated to maximizing the ratio between the classification loss on the reweighted dataset and the uncertainty of the ground-truth class labels. This is a minimax problem that REPAIR solves by alternatingly updating classifier parameters and dataset resampling weights, using stochastic gradient descent. An experimental set-up is also introduced to measure the bias of any dataset for a given representation, and the impact of this bias on the performance of recognition models. Experiments with synthetic and action recognition data show that dataset REPAIR can significantly reduce representation bias, and lead to improved generalization of models trained on REPAIRed datasets. The tools used for characterizing representation bias, and the proposed dataset REPAIR algorithm, are available at https://github.com/JerryYLi/Dataset-REPAIR/.

Zhaowei Cai, Nuno Vasconcelos

In object detection, an intersection over union (IoU) threshold is required to define positives and negatives. An object detector, trained with low IoU threshold, e.g. 0.5, usually produces noisy detections. However, detection performance tends to degrade with increasing the IoU thresholds. Two main factors are responsible for this: 1) overfitting during training, due to exponentially vanishing positive samples, and 2) inference-time mismatch between the IoUs for which the detector is optimal and those of the input hypotheses. A multi-stage object detection architecture, the Cascade R-CNN, is proposed to address these problems. It consists of a sequence of detectors trained with increasing IoU thresholds, to be sequentially more selective against close false positives. The detectors are trained stage by stage, leveraging the observation that the output of a detector is a good distribution for training the next higher quality detector. The resampling of progressively improved hypotheses guarantees that all detectors have a positive set of examples of equivalent size, reducing the overfitting problem. The same cascade procedure is applied at inference, enabling a closer match between the hypotheses and the detector quality of each stage. A simple implementation of the Cascade R-CNN is shown to surpass all single-model object detectors on the challenging COCO dataset. Experiments also show that the Cascade R-CNN is widely applicable across detector architectures, achieving consistent gains independently of the baseline detector strength. The code will be made available at https://github.com/zhaoweicai/cascade-rcnn.

Jiteng Mu, Nuno Vasconcelos, Xiaolong Wang

Recent progress in controllable image generation and editing is largely driven by diffusion-based methods. Although diffusion models perform exceptionally well in specific tasks with tailored designs, establishing a unified model is still challenging. In contrast, autoregressive models inherently feature a unified tokenized representation, which simplifies the creation of a single foundational model for various tasks. In this work, we propose EditAR, a single unified autoregressive framework for a variety of conditional image generation tasks, e.g., image editing, depth-to-image, edge-to-image, segmentation-to-image. The model takes both images and instructions as inputs, and predicts the edited images tokens in a vanilla next-token paradigm. To enhance the text-to-image alignment, we further propose to distill the knowledge from foundation models into the autoregressive modeling process. We evaluate its effectiveness across diverse tasks on established benchmarks, showing competitive performance to various state-of-the-art task-specific methods. Project page: https://jitengmu.github.io/EditAR/

Chih-Hui Ho, Kuan-Chuan Peng, Nuno Vasconcelos

Anomaly detection (AD) aims to identify defective images and localize their defects (if any). Ideally, AD models should be able to detect defects over many image classes; without relying on hard-coded class names that can be uninformative or inconsistent across datasets; learn without anomaly supervision; and be robust to the long-tailed distributions of real-world applications. To address these challenges, we formulate the problem of long-tailed AD by introducing several datasets with different levels of class imbalance and metrics for performance evaluation. We then propose a novel method, LTAD, to detect defects from multiple and long-tailed classes, without relying on dataset class names. LTAD combines AD by reconstruction and semantic AD modules. AD by reconstruction is implemented with a transformer-based reconstruction module. Semantic AD is implemented with a binary classifier, which relies on learned pseudo class names and a pretrained foundation model. These modules are learned over two phases. Phase 1 learns the pseudo-class names and a variational autoencoder (VAE) for feature synthesis that augments the training data to combat long-tails. Phase 2 then learns the parameters of the reconstruction and classification modules of LTAD. Extensive experiments using the proposed long-tailed datasets show that LTAD substantially outperforms the state-of-the-art methods for most forms of dataset imbalance. The long-tailed dataset split is available at https://zenodo.org/records/10854201 .

Jiteng Mu, Michaël Gharbi, Richard Zhang et al.

Diffusion models have made significant advances in text-guided synthesis tasks. However, editing user-provided images remains challenging, as the high dimensional noise input space of diffusion models is not naturally suited for image inversion or spatial editing. In this work, we propose an image representation that promotes spatial editing of input images using a diffusion model. Concretely, we learn to encode an input into "image elements" that can faithfully reconstruct an input image. These elements can be intuitively edited by a user, and are decoded by a diffusion model into realistic images. We show the effectiveness of our representation on various image editing tasks, such as object resizing, rearrangement, dragging, de-occlusion, removal, variation, and image composition. Project page: https://jitengmu.github.io/Editable_Image_Elements/

Alakh Desai, Nuno Vasconcelos

For image generation with diffusion models (DMs), a negative prompt n can be used to complement the text prompt p, helping define properties not desired in the synthesized image. While this improves prompt adherence and image quality, finding good negative prompts is challenging. We argue that this is due to a semantic gap between humans and DMs, which makes good negative prompts for DMs appear unintuitive to humans. To bridge this gap, we propose a new diffusion-negative prompting (DNP) strategy. DNP is based on a new procedure to sample images that are least compliant with p under the distribution of the DM, denoted as diffusion-negative sampling (DNS). Given p, one such image is sampled, which is then translated into natural language by the user or a captioning model, to produce the negative prompt n*. The pair (p, n*) is finally used to prompt the DM. DNS is straightforward to implement and requires no training. Experiments and human evaluations show that DNP performs well both quantitatively and qualitatively and can be easily combined with several DM variants.

Deepak Sridhar, Abhishek Peri, Rohith Rachala et al.

Recent advances in generative modeling with diffusion processes (DPs) enabled breakthroughs in image synthesis. Despite impressive image quality, these models have various prompt compliance problems, including low recall in generating multiple objects, difficulty in generating text in images, and meeting constraints like object locations and pose. For fine-grained editing and manipulation, they also require fine-grained semantic or instance maps that are tedious to produce manually. While prompt compliance can be enhanced by addition of loss functions at inference, this is time consuming and does not scale to complex scenes. To overcome these limitations, this work introduces a new family of \textit{Factor Graph Diffusion Models} (FG-DMs) that models the joint distribution of images and conditioning variables, such as semantic, sketch, depth or normal maps via a factor graph decomposition. This joint structure has several advantages, including support for efficient sampling based prompt compliance schemes, which produce images of high object recall, semi-automated fine-grained editing, text-based editing of conditions with noise inversion, explainability at intermediate levels, ability to produce labeled datasets for the training of downstream models such as segmentation or depth, training with missing data, and continual learning where new conditioning variables can be added with minimal or no modifications to the existing structure. We propose an implementation of FG-DMs by adapting a pre-trained Stable Diffusion (SD) model to implement all FG-DM factors, using only COCO dataset, and show that it is effective in generating images with 15\% higher recall than SD while retaining its generalization ability. We introduce an attention distillation loss that encourages consistency among the attention maps of all factors, improving the fidelity of the generated conditions and image.

Anand Kumar, Jiteng Mu, Nuno Vasconcelos

Text-to-image (T2I) models have recently gained widespread adoption. This has spurred concerns about safeguarding intellectual property rights and an increasing demand for mechanisms that prevent the generation of specific artistic styles. Existing methods for style extraction typically necessitate the collection of custom datasets and the training of specialized models. This, however, is resource-intensive, time-consuming, and often impractical for real-time applications. We present a novel, training-free framework to solve the style attribution problem, using the features produced by a diffusion model alone, without any external modules or retraining. This is denoted as Introspective Style attribution (IntroStyle) and is shown to have superior performance to state-of-the-art models for style attribution. We also introduce a synthetic dataset of Artistic Style Split (ArtSplit) to isolate artistic style and evaluate fine-grained style attribution performance. Our experimental results on WikiArt and DomainNet datasets show that \ours is robust to the dynamic nature of artistic styles, outperforming existing methods by a wide margin.

Zihao Zhao, Yijiang Li, Yuchen Yang et al.

Machine unlearning--enabling a trained model to forget specific data--is crucial for addressing biased data and adhering to privacy regulations like the General Data Protection Regulation (GDPR)'s "right to be forgotten". Recent works have paid little attention to privacy concerns, leaving the data intended for forgetting vulnerable to membership inference attacks. Moreover, they often come with high computational overhead. In this work, we propose Pseudo-Probability Unlearning (PPU), a novel method that enables models to forget data efficiently and in a privacy-preserving manner. Our method replaces the final-layer output probabilities of the neural network with pseudo-probabilities for the data to be forgotten. These pseudo-probabilities follow either a uniform distribution or align with the model's overall distribution, enhancing privacy and reducing risk of membership inference attacks. Our optimization strategy further refines the predictive probability distributions and updates the model's weights accordingly, ensuring effective forgetting with minimal impact on the model's overall performance. Through comprehensive experiments on multiple benchmarks, our method achieves over 20% improvements in forgetting error compared to the state-of-the-art. Additionally, our method enhances privacy by preventing the forgotten set from being inferred to around random guesses.

Deepak Sridhar, Kartikeya Bhardwaj, Jeya Pradha Jeyaraj et al.

Video reasoning using Large Multimodal Models (LMMs) relies on costly reinforcement learning (RL) and verbose chain-of-thought, resulting in substantial computational overhead during both training and inference. Moreover, the mechanisms that control the thinking process in these reasoning models are very limited. In this paper, using entropy of the model's output as a signal, we discover that the high-quality models go through a series of micro-explorations and micro-exploitations which keep the reasoning process grounded (i.e., avoid excessive randomness while the model is exploring or thinking through an answer). We further observe that once this "thinking" process is over, more accurate models demonstrate a better convergence by reducing the entropy significantly via a final exploitation phase (i.e., a more certain convergence towards a solution trajectory). We then use these novel, theoretically-grounded insights to tune the model's behavior directly at inference, without using any RL or supervised fine-tuning. Specifically, during inference, our proposed approach called V-Reason (Video-Reason) adapts the value cache of the LMM via a few optimization steps on a small, trainable controller using an entropy-based objective, i.e., no supervision from any dataset or RL is necessary. This tuning improves the model's micro-exploration and exploitation behavior during inference. Our experiments show that our proposed method achieves significant improvements over the base instruction-tuned models across several video reasoning datasets, narrowing the gap with RL-trained models to within 0.6% average accuracy without any training, while offering massive efficiency benefits: output tokens are reduced by 58.6% compared to the RL model.

Alakh Desai, Nuno Vasconcelos

Diffusion models (DMs) have demonstrated an unparalleled ability to create diverse and high-fidelity images from text prompts. However, they are also well-known to vary substantially regarding both prompt adherence and quality. Negative prompting was introduced to improve prompt compliance by specifying what an image must not contain. Previous works have shown the existence of an ideal negative prompt that can maximize the odds of the positive prompt. In this work, we explore relations between negative prompting and classifier-free guidance (CFG) to develop a sampling procedure, {\it Adaptive Negative Sampling Without External Resources} (ANSWER), that accounts for both positive and negative conditions from a single prompt. This leverages the internal understanding of negation by the diffusion model to increase the odds of generating images faithful to the prompt. ANSWER is a training-free technique, applicable to any model that supports CFG, and allows for negative grounding of image concepts without an explicit negative prompts, which are lossy and incomplete. Experiments show that adding ANSWER to existing DMs outperforms the baselines on multiple benchmarks and is preferred by humans 2x more over the other methods.

Jiacheng Cheng, Hijung Valentina Shin, Nuno Vasconcelos et al.

In the recent years, the dual-encoder vision-language models (\eg CLIP) have achieved remarkable text-to-image retrieval performance. However, we discover that these models usually results in very different retrievals for a pair of paraphrased queries. Such behavior might render the retrieval system less predictable and lead to user frustration. In this work, we consider the task of paraphrased text-to-image retrieval where a model aims to return similar results given a pair of paraphrased queries. To start with, we collect a dataset of paraphrased image descriptions to facilitate quantitative evaluation for this task. We then hypothesize that the undesired behavior of existing dual-encoder model is due to their text towers which are trained on image-sentence pairs and lack the ability to capture the semantic similarity between paraphrased queries. To improve on this, we investigate multiple strategies for training a dual-encoder model starting from a language model pretrained on a large text corpus. Compared to public dual-encoder models such as CLIP and OpenCLIP, the model trained with our best adaptation strategy achieves a significantly higher ranking similarity for paraphrased queries while maintaining similar zero-shot classification and retrieval accuracy.

Zhen Wang, Yuelei Li, Jia Wan et al.

Crowd counting is an important problem in computer vision due to its wide range of applications in image understanding. Currently, this problem is typically addressed using deep learning approaches, such as Convolutional Neural Networks (CNNs) and Transformers. However, deep networks are data-driven and are prone to overfitting, especially when the available labeled crowd dataset is limited. To overcome this limitation, we have designed a pipeline that utilizes a diffusion model to generate extensive training data. We are the first to generate images conditioned on a location dot map (a binary dot map that specifies the location of human heads) with a diffusion model. We are also the first to use these diverse synthetic data to augment the crowd counting models. Our proposed smoothed density map input for ControlNet significantly improves ControlNet's performance in generating crowds in the correct locations. Also, Our proposed counting loss for the diffusion model effectively minimizes the discrepancies between the location dot map and the crowd images generated. Additionally, our innovative guidance sampling further directs the diffusion process toward regions where the generated crowd images align most accurately with the location dot map. Collectively, we have enhanced ControlNet's ability to generate specified objects from a location dot map, which can be used for data augmentation in various counting problems. Moreover, our framework is versatile and can be easily adapted to all kinds of counting problems. Extensive experiments demonstrate that our framework improves the counting performance on the ShanghaiTech, NWPU-Crowd, UCF-QNRF, and TRANCOS datasets, showcasing its effectiveness.

Jiteng Mu, Shalini De Mello, Zhiding Yu et al.

Recent advances show that Generative Adversarial Networks (GANs) can synthesize images with smooth variations along semantically meaningful latent directions, such as pose, expression, layout, etc. While this indicates that GANs implicitly learn pixel-level correspondences across images, few studies explored how to extract them explicitly. In this work, we introduce Coordinate GAN (CoordGAN), a structure-texture disentangled GAN that learns a dense correspondence map for each generated image. We represent the correspondence maps of different images as warped coordinate frames transformed from a canonical coordinate frame, i.e., the correspondence map, which describes the structure (e.g., the shape of a face), is controlled via a transformation. Hence, finding correspondences boils down to locating the same coordinate in different correspondence maps. In CoordGAN, we sample a transformation to represent the structure of a synthesized instance, while an independent texture branch is responsible for rendering appearance details orthogonal to the structure. Our approach can also extract dense correspondence maps for real images by adding an encoder on top of the generator. We quantitatively demonstrate the quality of the learned dense correspondences through segmentation mask transfer on multiple datasets. We also show that the proposed generator achieves better structure and texture disentanglement compared to existing approaches. Project page: https://jitengmu.github.io/CoordGAN/

Zhirui Dai, Yuepeng Jiang, Yi Li et al.

Social distancing, an essential public health measure to limit the spread of contagious diseases, has gained significant attention since the outbreak of the COVID-19 pandemic. In this work, the problem of visual social distancing compliance assessment in busy public areas, with wide field-of-view cameras, is considered. A dataset of crowd scenes with people annotations under a bird's eye view (BEV) and ground truth for metric distances is introduced, and several measures for the evaluation of social distance detection systems are proposed. A multi-branch network, BEV-Net, is proposed to localize individuals in world coordinates and identify high-risk regions where social distancing is violated. BEV-Net combines detection of head and feet locations, camera pose estimation, a differentiable homography module to map image into BEV coordinates, and geometric reasoning to produce a BEV map of the people locations in the scene. Experiments on complex crowded scenes demonstrate the power of the approach and show superior performance over baselines derived from methods in the literature. Applications of interest for public health decision makers are finally discussed. Datasets, code and pretrained models are publicly available at GitHub.

Brandon Leung, Chih-Hui Ho, Amir Persekian et al.

The hypothesis that image datasets gathered online "in the wild" can produce biased object recognizers, e.g. preferring professional photography or certain viewing angles, is studied. A new "in the lab" data collection infrastructure is proposed consisting of a drone which captures images as it circles around objects. Crucially, the control provided by this setup and the natural camera shake inherent to flight mitigate many biases. It's inexpensive and easily replicable nature may also potentially lead to a scalable data collection effort by the vision community. The procedure's usefulness is demonstrated by creating a dataset of Objects Obtained With fLight (OOWL). Denoted as OOWL500, it contains 120,000 images of 500 objects and is the largest "in the lab" image dataset available when both number of classes and objects per class are considered. Furthermore, it has enabled several of new insights on object recognition. First, a novel adversarial attack strategy is proposed, where image perturbations are defined in terms of semantic properties such as camera shake and pose. Indeed, experiments have shown that ImageNet has considerable amounts of pose and professional photography bias. Second, it is used to show that the augmentation of in the wild datasets, such as ImageNet, with in the lab data, such as OOWL500, can significantly decrease these biases, leading to object recognizers of improved generalization. Third, the dataset is used to study questions on "best procedures" for dataset collection. It is revealed that data augmentation with synthetic images does not suffice to eliminate in the wild datasets biases, and that camera shake and pose diversity play a more important role in object recognition robustness than previously thought.

Brandon Leung, Chih-Hui Ho, Nuno Vasconcelos

Much recent progress has been made in reconstructing the 3D shape of an object from an image of it, i.e. single view 3D reconstruction. However, it has been suggested that current methods simply adopt a "nearest-neighbor" strategy, instead of genuinely understanding the shape behind the input image. In this paper, we rigorously show that for many state of the art methods, this issue manifests as (1) inconsistencies between coarse reconstructions and input images, and (2) inability to generalize across domains. We thus propose REFINE, a postprocessing mesh refinement step that can be easily integrated into the pipeline of any black-box method in the literature. At test time, REFINE optimizes a network per mesh instance, to encourage consistency between the mesh and the given object view. This, along with a novel combination of regularizing losses, reduces the domain gap and achieves state of the art performance. We believe that this novel paradigm is an important step towards robust, accurate reconstructions, remaining relevant as new reconstruction networks are introduced.

Alakh Desai, Tz-Ying Wu, Subarna Tripathi et al.

Significant effort has been recently devoted to modeling visual relations. This has mostly addressed the design of architectures, typically by adding parameters and increasing model complexity. However, visual relation learning is a long-tailed problem, due to the combinatorial nature of joint reasoning about groups of objects. Increasing model complexity is, in general, ill-suited for long-tailed problems due to their tendency to overfit. In this paper, we explore an alternative hypothesis, denoted the Devil is in the Tails. Under this hypothesis, better performance is achieved by keeping the model simple but improving its ability to cope with long-tailed distributions. To test this hypothesis, we devise a new approach for training visual relationships models, which is inspired by state-of-the-art long-tailed recognition literature. This is based on an iterative decoupled training scheme, denoted Decoupled Training for Devil in the Tails (DT2). DT2 employs a novel sampling approach, Alternating Class-Balanced Sampling (ACBS), to capture the interplay between the long-tailed entity and predicate distributions of visual relations. Results show that, with an extremely simple architecture, DT2-ACBS significantly outperforms much more complex state-of-the-art methods on scene graph generation tasks. This suggests that the development of sophisticated models must be considered in tandem with the long-tailed nature of the problem.

Bo Liu, Haoxiang Li, Hao Kang et al.

Main challenges in long-tailed recognition come from the imbalanced data distribution and sample scarcity in its tail classes. While techniques have been proposed to achieve a more balanced training loss and to improve tail classes data variations with synthesized samples, we resort to leverage readily available unlabeled data to boost recognition accuracy. The idea leads to a new recognition setting, namely semi-supervised long-tailed recognition. We argue this setting better resembles the real-world data collection and annotation process and hence can help close the gap to real-world scenarios. To address the semi-supervised long-tailed recognition problem, we present an alternate sampling framework combining the intuitions from successful methods in these two research areas. The classifier and feature embedding are learned separately and updated iteratively. The class-balanced sampling strategy has been implemented to train the classifier in a way not affected by the pseudo labels' quality on the unlabeled data. A consistency loss has been introduced to limit the impact from unlabeled data while leveraging them to update the feature embedding. We demonstrate significant accuracy improvements over other competitive methods on two datasets.

Bo Liu, Haoxiang Li, Hao Kang et al.

The problem of long-tailed recognition, where the number of examples per class is highly unbalanced, is considered. It is hypothesized that the well known tendency of standard classifier training to overfit to popular classes can be exploited for effective transfer learning. Rather than eliminating this overfitting, e.g. by adopting popular class-balanced sampling methods, the learning algorithm should instead leverage this overfitting to transfer geometric information from popular to low-shot classes. A new classifier architecture, GistNet, is proposed to support this goal, using constellations of classifier parameters to encode the class geometry. A new learning algorithm is then proposed for GeometrIc Structure Transfer (GIST), with resort to a combination of loss functions that combine class-balanced and random sampling to guarantee that, while overfitting to the popular classes is restricted to geometric parameters, it is leveraged to transfer class geometry from popular to few-shot classes. This enables better generalization for few-shot classes without the need for the manual specification of class weights, or even the explicit grouping of classes into different types. Experiments on two popular long-tailed recognition datasets show that GistNet outperforms existing solutions to this problem.

Bo Liu, Haoxiang Li, Hao Kang et al.

The problem of long-tailed recognition, where the number of examples per class is highly unbalanced, is considered. While training with class-balanced sampling has been shown effective for this problem, it is known to over-fit to few-shot classes. It is hypothesized that this is due to the repeated sampling of examples and can be addressed by feature space augmentation. A new feature augmentation strategy, EMANATE, based on back-tracking of features across epochs during training, is proposed. It is shown that, unlike class-balanced sampling, this is an adversarial augmentation strategy. A new sampling procedure, Breadcrumb, is then introduced to implement adversarial class-balanced sampling without extra computation. Experiments on three popular long-tailed recognition datasets show that Breadcrumb training produces classifiers that outperform existing solutions to the problem.

Bo Liu, Mandar Dixit, Roland Kwitt et al.

We propose a method to learn, even using a dataset where objects appear only in sparsely sampled views (e.g. Pix3D), the ability to synthesize a pose trajectory for an arbitrary reference image. This is achieved with a cross-modal pose trajectory transfer mechanism. First, a domain transfer function is trained to predict, from an RGB image of the object, its 2D depth map. Then, a set of image views is generated by learning to simulate object rotation in the depth space. Finally, the generated poses are mapped from this latent space into a set of corresponding RGB images using a learned identity preserving transform. This results in a dense pose trajectory of the object in image space. For each object type (e.g., a specific Ikea chair model), a 3D CAD model is used to render a full pose trajectory of 2D depth maps. In the absence of dense pose sampling in image space, these latent space trajectories provide cross-modal guidance for learning. The learned pose trajectories can be transferred to unseen examples, effectively synthesizing all object views in image space. Our method is evaluated on the Pix3D and ShapeNet datasets, in the setting of novel view synthesis under sparse pose supervision, demonstrating substantial improvements over recent art.