Lorenzo Torresani

Kristen Grauman, Andrew Westbury, Lorenzo Torresani et al. · cmu, gatech

We present Ego-Exo4D, a diverse, large-scale multimodal multiview video dataset and benchmark challenge. Ego-Exo4D centers around simultaneously-captured egocentric and exocentric video of skilled human activities (e.g., sports, music, dance, bike repair). 740 participants from 13 cities worldwide performed these activities in 123 different natural scene contexts, yielding long-form captures from 1 to 42 minutes each and 1,286 hours of video combined. The multimodal nature of the dataset is unprecedented: the video is accompanied by multichannel audio, eye gaze, 3D point clouds, camera poses, IMU, and multiple paired language descriptions -- including a novel "expert commentary" done by coaches and teachers and tailored to the skilled-activity domain. To push the frontier of first-person video understanding of skilled human activity, we also present a suite of benchmark tasks and their annotations, including fine-grained activity understanding, proficiency estimation, cross-view translation, and 3D hand/body pose. All resources are open sourced to fuel new research in the community. Project page: http://ego-exo4d-data.org/

Kumar Ashutosh, Rohit Girdhar, Lorenzo Torresani et al. · meta-ai

Video-language embeddings are a promising avenue for injecting semantics into visual representations, but existing methods capture only short-term associations between seconds-long video clips and their accompanying text. We propose HierVL, a novel hierarchical video-language embedding that simultaneously accounts for both long-term and short-term associations. As training data, we take videos accompanied by timestamped text descriptions of human actions, together with a high-level text summary of the activity throughout the long video (as are available in Ego4D). We introduce a hierarchical contrastive training objective that encourages text-visual alignment at both the clip level and video level. While the clip-level constraints use the step-by-step descriptions to capture what is happening in that instant, the video-level constraints use the summary text to capture why it is happening, i.e., the broader context for the activity and the intent of the actor. Our hierarchical scheme yields a clip representation that outperforms its single-level counterpart as well as a long-term video representation that achieves SotA results on tasks requiring long-term video modeling. HierVL successfully transfers to multiple challenging downstream tasks (in EPIC-KITCHENS-100, Charades-Ego, HowTo100M) in both zero-shot and fine-tuned settings.

Kumar Ashutosh, Rohit Girdhar, Lorenzo Torresani et al. · meta-ai

Narrated ''how-to'' videos have emerged as a promising data source for a wide range of learning problems, from learning visual representations to training robot policies. However, this data is extremely noisy, as the narrations do not always describe the actions demonstrated in the video. To address this problem we introduce the novel task of visual narration detection, which entails determining whether a narration is visually depicted by the actions in the video. We propose What You Say is What You Show (WYS^2), a method that leverages multi-modal cues and pseudo-labeling to learn to detect visual narrations with only weakly labeled data. Our model successfully detects visual narrations in in-the-wild videos, outperforming strong baselines, and we demonstrate its impact for state-of-the-art summarization and temporal alignment of instructional videos.

Raghav Goyal, Effrosyni Mavroudi, Xitong Yang et al. · meta-ai

Video understanding tasks take many forms, from action detection to visual query localization and spatio-temporal grounding of sentences. These tasks differ in the type of inputs (only video, or video-query pair where query is an image region or sentence) and outputs (temporal segments or spatio-temporal tubes). However, at their core they require the same fundamental understanding of the video, i.e., the actors and objects in it, their actions and interactions. So far these tasks have been tackled in isolation with individual, highly specialized architectures, which do not exploit the interplay between tasks. In contrast, in this paper, we present a single, unified model for tackling query-based video understanding in long-form videos. In particular, our model can address all three tasks of the Ego4D Episodic Memory benchmark which entail queries of three different forms: given an egocentric video and a visual, textual or activity query, the goal is to determine when and where the answer can be seen within the video. Our model design is inspired by recent query-based approaches to spatio-temporal grounding, and contains modality-specific query encoders and task-specific sliding window inference that allow multi-task training with diverse input modalities and different structured outputs. We exhaustively analyze relationships among the tasks and illustrate that cross-task learning leads to improved performance on each individual task, as well as the ability to generalize to unseen tasks, such as zero-shot spatial localization of language queries.

Zihui Xue, Yale Song, Kristen Grauman et al.

Different video understanding tasks are typically treated in isolation, and even with distinct types of curated data (e.g., classifying sports in one dataset, tracking animals in another). However, in wearable cameras, the immersive egocentric perspective of a person engaging with the world around them presents an interconnected web of video understanding tasks -- hand-object manipulations, navigation in the space, or human-human interactions -- that unfold continuously, driven by the person's goals. We argue that this calls for a much more unified approach. We propose EgoTask Translation (EgoT2), which takes a collection of models optimized on separate tasks and learns to translate their outputs for improved performance on any or all of them at once. Unlike traditional transfer or multi-task learning, EgoT2's flipped design entails separate task-specific backbones and a task translator shared across all tasks, which captures synergies between even heterogeneous tasks and mitigates task competition. Demonstrating our model on a wide array of video tasks from Ego4D, we show its advantages over existing transfer paradigms and achieve top-ranked results on four of the Ego4D 2022 benchmark challenges.

Huiyu Wang, Mitesh Kumar Singh, Lorenzo Torresani

We present Ego-Only, the first approach that enables state-of-the-art action detection on egocentric (first-person) videos without any form of exocentric (third-person) transferring. Despite the content and appearance gap separating the two domains, large-scale exocentric transferring has been the default choice for egocentric action detection. This is because prior works found that egocentric models are difficult to train from scratch and that transferring from exocentric representations leads to improved accuracy. However, in this paper, we revisit this common belief. Motivated by the large gap separating the two domains, we propose a strategy that enables effective training of egocentric models without exocentric transferring. Our Ego-Only approach is simple. It trains the video representation with a masked autoencoder finetuned for temporal segmentation. The learned features are then fed to an off-the-shelf temporal action localization method to detect actions. We find that this renders exocentric transferring unnecessary by showing remarkably strong results achieved by this simple Ego-Only approach on three established egocentric video datasets: Ego4D, EPIC-Kitchens-100, and Charades-Ego. On both action detection and action recognition, Ego-Only outperforms previous best exocentric transferring methods that use orders of magnitude more labels. Ego-Only sets new state-of-the-art results on these datasets and benchmarks without exocentric data.

Effrosyni Mavroudi, Triantafyllos Afouras, Lorenzo Torresani

In this paper we present an approach for localizing steps of procedural activities in narrated how-to videos. To deal with the scarcity of labeled data at scale, we source the step descriptions from a language knowledge base (wikiHow) containing instructional articles for a large variety of procedural tasks. Without any form of manual supervision, our model learns to temporally ground the steps of procedural articles in how-to videos by matching three modalities: frames, narrations, and step descriptions. Specifically, our method aligns steps to video by fusing information from two distinct pathways: i) {\em direct} alignment of step descriptions to frames, ii) {\em indirect} alignment obtained by composing steps-to-narrations with narrations-to-video correspondences. Notably, our approach performs global temporal grounding of all steps in an article at once by exploiting order information, and is trained with step pseudo-labels which are iteratively refined and aggressively filtered. In order to validate our model we introduce a new evaluation benchmark -- HT-Step -- obtained by manually annotating a 124-hour subset of HowTo100M\footnote{A test server is accessible at \url{https://eval.ai/web/challenges/challenge-page/2082}.} with steps sourced from wikiHow articles. Experiments on this benchmark as well as zero-shot evaluations on CrossTask demonstrate that our multi-modality alignment yields dramatic gains over several baselines and prior works. Finally, we show that our inner module for matching narration-to-video outperforms by a large margin the state of the art on the HTM-Align narration-video alignment benchmark.

Reuben Tan, Matthias De Lange, Michael Iuzzolino et al.

Long-term activity forecasting is an especially challenging research problem because it requires understanding the temporal relationships between observed actions, as well as the variability and complexity of human activities. Despite relying on strong supervision via expensive human annotations, state-of-the-art forecasting approaches often generalize poorly to unseen data. To alleviate this issue, we propose Multiscale Video Pretraining (MVP), a novel self-supervised pretraining approach that learns robust representations for forecasting by learning to predict contextualized representations of future video clips over multiple timescales. MVP is based on our observation that actions in videos have a multiscale nature, where atomic actions typically occur at a short timescale and more complex actions may span longer timescales. We compare MVP to state-of-the-art self-supervised video learning approaches on downstream long-term forecasting tasks including long-term action anticipation and video summary prediction. Our comprehensive experiments across the Ego4D and Epic-Kitchens-55/100 datasets demonstrate that MVP out-performs state-of-the-art methods by significant margins. Notably, MVP obtains a relative performance gain of over 20% accuracy in video summary forecasting over existing methods.

Zihui Xue, Yale Song, Kristen Grauman et al.

This technical report describes the EgoTask Translation approach that explores relations among a set of egocentric video tasks in the Ego4D challenge. To improve the primary task of interest, we propose to leverage existing models developed for other related tasks and design a task translator that learns to ''translate'' auxiliary task features to the primary task. With no modification to the baseline architectures, our proposed approach achieves competitive performance on two Ego4D challenges, ranking the 1st in the talking to me challenge and the 3rd in the PNR keyframe localization challenge.

Tarun Kalluri, Weiyao Wang, Heng Wang et al.

Many top-down architectures for instance segmentation achieve significant success when trained and tested on pre-defined closed-world taxonomy. However, when deployed in the open world, they exhibit notable bias towards seen classes and suffer from significant performance drop. In this work, we propose a novel approach for open world instance segmentation called bottom-Up and top-Down Open-world Segmentation (UDOS) that combines classical bottom-up segmentation algorithms within a top-down learning framework. UDOS first predicts parts of objects using a top-down network trained with weak supervision from bottom-up segmentations. The bottom-up segmentations are class-agnostic and do not overfit to specific taxonomies. The part-masks are then fed into affinity-based grouping and refinement modules to predict robust instance-level segmentations. UDOS enjoys both the speed and efficiency from the top-down architectures and the generalization ability to unseen categories from bottom-up supervision. We validate the strengths of UDOS on multiple cross-category as well as cross-dataset transfer tasks from 5 challenging datasets including MS-COCO, LVIS, ADE20k, UVO and OpenImages, achieving significant improvements over state-of-the-art across the board. Our code and models are available on our project page.

Joseph DiPalma, Lorenzo Torresani, Saeed Hassanpour

Deep learning has been effective for histology image analysis in digital pathology. However, many current deep learning approaches require large, strongly- or weakly-labeled images and regions of interest, which can be time-consuming and resource-intensive to obtain. To address this challenge, we present HistoPerm, a view generation method for representation learning using joint embedding architectures that enhances representation learning for histology images. HistoPerm permutes augmented views of patches extracted from whole-slide histology images to improve classification performance. We evaluated the effectiveness of HistoPerm on two histology image datasets for Celiac disease and Renal Cell Carcinoma, using three widely used joint embedding architecture-based representation learning methods: BYOL, SimCLR, and VICReg. Our results show that HistoPerm consistently improves patch- and slide-level classification performance in terms of accuracy, F1-score, and AUC. Specifically, for patch-level classification accuracy on the Celiac disease dataset, HistoPerm boosts BYOL and VICReg by 8% and SimCLR by 3%. On the Renal Cell Carcinoma dataset, patch-level classification accuracy is increased by 2% for BYOL and VICReg, and by 1% for SimCLR. In addition, on the Celiac disease dataset, models with HistoPerm outperform the fully-supervised baseline model by 6%, 5%, and 2% for BYOL, SimCLR, and VICReg, respectively. For the Renal Cell Carcinoma dataset, HistoPerm lowers the classification accuracy gap for the models up to 10% relative to the fully-supervised baseline. These findings suggest that HistoPerm can be a valuable tool for improving representation learning of histopathology features when access to labeled data is limited and can lead to whole-slide classification results that are comparable to or superior to fully-supervised methods.

Yulu Pan, Han Yi, Seongsu Ha et al.

True video intelligence demands more than recognizing what is visible: it requires reasoning about why events unfold, predicting what would change under different conditions, and deciding what to do next. We refer to this progression, from perception through causal reasoning and simulation to strategic planning, as Strategic Video Intelligence (SVI). No existing benchmark evaluates this capability stack: in-the-wild videos lack verifiable ground truth for causal and strategic questions, while synthetic environments sacrifice the complexity of real multi-agent systems. To bridge this gap, we introduce SVI-Bench, a large-scale benchmark that leverages team sports as a dynamic microworld, combining the complexity of real-world multi-agent interaction (10-22 agents making coordinated decisions under adversarial pressure) with the verifiability of explicit rules and definitive outcomes. SVI-Bench comprises approximately 35K hours of broadcast video, 15M annotated actions, 15K hours of expert commentary, 23K game reports, and 103K structured statistical records across basketball, soccer, and hockey, all constructed via a data engine that transforms raw game data into a dense, cross-referenced corpus. We organize evaluation into 9 tasks spanning a progressive four-pillar hierarchy: Dynamic Scene Understanding, Causal Reasoning, Strategic Simulation, and Agentic Synthesis. Evaluating strong multimodal and agentic baselines, we find a capability cliff: models perform competently on perceptual tasks, achieving approximately 73% on fine-grained action QA, but degrade sharply at each successive cognitive level. Agentic tasks prove hardest: the strongest model achieves only 5% accuracy when required to autonomously gather and integrate evidence across a corpus of 1.8M clips.

Maxwell Aladago, Lorenzo Torresani, Soroush Vosoughi

In the field of vision-language contrastive learning, models such as CLIP capitalize on matched image-caption pairs as positive examples and leverage within-batch non-matching pairs as negatives. This approach has led to remarkable outcomes in zero-shot image classification, cross-modal retrieval, and linear evaluation tasks. We show that the zero-shot classification and retrieval capabilities of CLIP-like models can be improved significantly through the introduction of semantically composite examples during pretraining. Inspired by CutMix in vision categorization, we create semantically composite image-caption pairs by merging elements from two distinct instances in the dataset via a novel procedure. Our method fuses the captions and blends 50% of each image to form a new composite sample. This simple technique (termed CLIP-C for CLIP Compositions), devoid of any additional computational overhead or increase in model parameters, significantly improves zero-shot image classification and cross-modal retrieval. The benefits of CLIP-C are particularly pronounced in settings with relatively limited pretraining data.

Jang Hyun Cho, Andrea Madotto, Effrosyni Mavroudi et al.

Vision-language models are integral to computer vision research, yet many high-performing models remain closed-source, obscuring their data, design and training recipe. The research community has responded by using distillation from black-box models to label training data, achieving strong benchmark results, at the cost of measurable scientific progress. However, without knowing the details of the teacher model and its data sources, scientific progress remains difficult to measure. In this paper, we study building a Perception Language Model (PLM) in a fully open and reproducible framework for transparent research in image and video understanding. We analyze standard training pipelines without distillation from proprietary models and explore large-scale synthetic data to identify critical data gaps, particularly in detailed video understanding. To bridge these gaps, we release 2.8M human-labeled instances of fine-grained video question-answer pairs and spatio-temporally grounded video captions. Additionally, we introduce PLM-VideoBench, a suite for evaluating challenging video understanding tasks focusing on the ability to reason about "what", "where", "when", and "how" of a video. We make our work fully reproducible by providing data, training recipes, code & models. https://github.com/facebookresearch/perception_models

Luigi Seminara, Antonino Furnari, Lorenzo Torresani

Visual planning asks a model to generate the remaining steps of a procedure in natural language given a partial video context and a goal. Progress on this task is bottlenecked by annotation: clean labeled datasets are small, domain-narrow, and encode a single execution trajectory per example, even though many valid orderings exist. Large-scale instructional video corpora offer orders of magnitude more procedural content, but supervised fine-tuning on pseudo-labels from their noisy ASR narrations propagates segmentation and alignment errors and stays single-trajectory. We identify a key asymmetry: extracting clean step labels from noisy video is hard, but verifying whether a generated step sequence is temporally grounded in ASR transcripts is cheap and scales to millions of videos via precomputed text embeddings. We exploit this asymmetry in RECIPE, which uses grounding quality as a reward for GRPO, turning the noisy corpus into a verifier rather than a label source. The framework applies uniformly to two planner input configurations (Socratic, with a textual history extracted by a frozen VLM, and Video, consuming video tokens directly) and to annotated and weakly supervised regimes. We evaluate on 7 procedural benchmarks using a reference-based LLM-as-judge protocol scoring plans across 6 procedural criteria. RECIPE-RL improves over the base checkpoint at all scales (0.5B, 3B, 7B) and every benchmark, with macro-accuracy gains of +7 to +8 points in-domain and up to +16 points zero-shot. It outperforms supervised fine-tuning on both annotated and pseudo-labeled plans (the latter degrades the base) and remains robust without human annotations. Used as the proposal stage of a prior propose-assess-search planner, it improves over the strongest zero-shot baseline at every horizon on Visual Planning for Assistance, and on COIN it preserves the generation diversity that SFT collapses.

Sejoon Jun, Hai Nguyen-Truong, Luigi Seminara et al.

Predicting how a person's first-person view will evolve (what action will follow, what plan completes a task, whether an in-progress shot will score) is fundamentally under-specified: the same context admits many plausible futures, and a model trained to minimize prediction error is forced to hedge or average across them, getting it wrong either way. Two findings shape our approach. First, the future camera trajectory, the path the head carves through space, lets the model commit to one of those futures: it carries the operator's intent in a form fine enough to determine how an action will unfold, substantially outperforming language as a conditioning signal. Second, this same intent makes the trajectory itself partially predictable from the context at hand, enough that trajectory need not be observed at test time to recover most of the gain. We instantiate these findings as TrajPilot, a model that predicts candidate future trajectories from egocentric context and uses them to pilot action prediction in an action-aligned embedding space where language shapes the structure but is never used as a conditioning input. TrajPilot beats VLM and structured-planner baselines on procedural planning across Ego-Exo4D atomic, Ego-Exo4D Keystep, Ego4D GoalStep, and EgoPER, with the trajectory advantage widening with horizon (exactly where prior planners collapse) and holding under RGB-only camera-pose estimation. With the goal masked at inference, the same model performs goal-free anticipation, beating VLM baselines on Ego-Exo4D atomic and extending to EPIC-Kitchens-100 and basketball shot-outcome prediction.

Tushar Nagarajan, Lorenzo Torresani

Comparing a user video to a reference how-to video is a key requirement for AR/VR technology delivering personalized assistance tailored to the user's progress. However, current approaches for language-based assistance can only answer questions about a single video. We propose an approach that first automatically generates large amounts of visual instruction tuning data involving pairs of videos from HowTo100M by leveraging existing step annotations and accompanying narrations, and then trains a video-conditioned language model to jointly reason across multiple raw videos. Our model achieves state-of-the-art performance at identifying differences between video pairs and ranking videos based on the severity of these differences, and shows promising ability to perform general reasoning over multiple videos. Project page: https://github.com/facebookresearch/stepdiff

Minjoon Jung, Byoung-Tak Zhang, Lorenzo Torresani

Video temporal grounding (VTG) takes an untrimmed video and a natural-language query as input and localizes the temporal moment that best matches the query. Existing methods rely on large, task-specific datasets requiring costly manual annotation. We introduce EvoGround, a framework of two coupled self-evolving agents, a proposer and a solver, that learn temporal grounding from raw videos without any human-labeled data. The proposer generates query--moment pairs from raw videos, while the solver learns to ground them and feeds back signals that improve the proposer in return. Through this self-reinforcing reinforcement-learning loop, the two agents are initialized from the same backbone and mutually improve across iterations. Trained on 2.5K unlabeled videos, EvoGround matches or surpasses fully supervised models across multiple VTG benchmarks, while emerging as a state-of-the-art fine-grained video captioner without manual labels.

Gedas Bertasius, Heng Wang, Lorenzo Torresani

We present a convolution-free approach to video classification built exclusively on self-attention over space and time. Our method, named "TimeSformer," adapts the standard Transformer architecture to video by enabling spatiotemporal feature learning directly from a sequence of frame-level patches. Our experimental study compares different self-attention schemes and suggests that "divided attention," where temporal attention and spatial attention are separately applied within each block, leads to the best video classification accuracy among the design choices considered. Despite the radically new design, TimeSformer achieves state-of-the-art results on several action recognition benchmarks, including the best reported accuracy on Kinetics-400 and Kinetics-600. Finally, compared to 3D convolutional networks, our model is faster to train, it can achieve dramatically higher test efficiency (at a small drop in accuracy), and it can also be applied to much longer video clips (over one minute long). Code and models are available at: https://github.com/facebookresearch/TimeSformer.

Gedas Bertasius, Christoph Feichtenhofer, Du Tran et al.

Modern approaches for multi-person pose estimation in video require large amounts of dense annotations. However, labeling every frame in a video is costly and labor intensive. To reduce the need for dense annotations, we propose a PoseWarper network that leverages training videos with sparse annotations (every k frames) to learn to perform dense temporal pose propagation and estimation. Given a pair of video frames---a labeled Frame A and an unlabeled Frame B---we train our model to predict human pose in Frame A using the features from Frame B by means of deformable convolutions to implicitly learn the pose warping between A and B. We demonstrate that we can leverage our trained PoseWarper for several applications. First, at inference time we can reverse the application direction of our network in order to propagate pose information from manually annotated frames to unlabeled frames. This makes it possible to generate pose annotations for the entire video given only a few manually-labeled frames. Compared to modern label propagation methods based on optical flow, our warping mechanism is much more compact (6M vs 39M parameters), and also more accurate (88.7% mAP vs 83.8% mAP). We also show that we can improve the accuracy of a pose estimator by training it on an augmented dataset obtained by adding our propagated poses to the original manual labels. Lastly, we can use our PoseWarper to aggregate temporal pose information from neighboring frames during inference. This allows our system to achieve state-of-the-art pose detection results on the PoseTrack2017 and PoseTrack2018 datasets. Code has been made available at: https://github.com/facebookresearch/PoseWarper.

Md Mohaiminul Islam, Ngan Ho, Xitong Yang et al.

Most video captioning models are designed to process short video clips of few seconds and output text describing low-level visual concepts (e.g., objects, scenes, atomic actions). However, most real-world videos last for minutes or hours and have a complex hierarchical structure spanning different temporal granularities. We propose Video ReCap, a recursive video captioning model that can process video inputs of dramatically different lengths (from 1 second to 2 hours) and output video captions at multiple hierarchy levels. The recursive video-language architecture exploits the synergy between different video hierarchies and can process hour-long videos efficiently. We utilize a curriculum learning training scheme to learn the hierarchical structure of videos, starting from clip-level captions describing atomic actions, then focusing on segment-level descriptions, and concluding with generating summaries for hour-long videos. Furthermore, we introduce Ego4D-HCap dataset by augmenting Ego4D with 8,267 manually collected long-range video summaries. Our recursive model can flexibly generate captions at different hierarchy levels while also being useful for other complex video understanding tasks, such as VideoQA on EgoSchema. Data, code, and models are available at: https://sites.google.com/view/vidrecap

Md Mohaiminul Islam, Tushar Nagarajan, Huiyu Wang et al.

Video Question Answering (VQA) in long videos poses the key challenge of extracting relevant information and modeling long-range dependencies from many redundant frames. The self-attention mechanism provides a general solution for sequence modeling, but it has a prohibitive cost when applied to a massive number of spatiotemporal tokens in long videos. Most prior methods rely on compression strategies to lower the computational cost, such as reducing the input length via sparse frame sampling or compressing the output sequence passed to the large language model (LLM) via space-time pooling. However, these naive approaches over-represent redundant information and often miss salient events or fast-occurring space-time patterns. In this work, we introduce BIMBA, an efficient state-space model to handle long-form videos. Our model leverages the selective scan algorithm to learn to effectively select critical information from high-dimensional video and transform it into a reduced token sequence for efficient LLM processing. Extensive experiments demonstrate that BIMBA achieves state-of-the-art accuracy on multiple long-form VQA benchmarks, including PerceptionTest, NExT-QA, EgoSchema, VNBench, LongVideoBench, and Video-MME. Code, and models are publicly available at https://sites.google.com/view/bimba-mllm.

Xizi Wang, Feng Cheng, Ziyang Wang et al.

Video temporal grounding aims to localize relevant temporal boundaries in a video given a textual prompt. Recent work has focused on enabling Video LLMs to perform video temporal grounding via next-token prediction of temporal timestamps. However, accurately localizing timestamps in videos remains challenging for Video LLMs when relying solely on temporal token prediction. Our proposed TimeRefine addresses this challenge in two ways. First, instead of directly predicting the start and end timestamps, we reformulate the temporal grounding task as a temporal refining task: the model first makes rough predictions and then refines them by predicting offsets to the target segment. This refining process is repeated multiple times, through which the model progressively self-improves its temporal localization accuracy. Second, to enhance the model's temporal perception capabilities, we incorporate an auxiliary prediction head that penalizes the model more if a predicted segment deviates further from the ground truth, thus encouraging the model to make closer and more accurate predictions. Our plug-and-play method can be integrated into most LLM-based temporal grounding approaches. The experimental results demonstrate that TimeRefine achieves 3.6% and 5.0% mIoU improvements on the ActivityNet and Charades-STA datasets, respectively. Code and pretrained models will be released.

Yujie Lu, Yale Song, William Wang et al.

We investigate complex video question answering via chain-of-evidence reasoning -- identifying sequences of temporal spans from multiple relevant parts of the video, together with visual evidence within them. Existing models struggle with multi-step reasoning as they uniformly sample a fixed number of frames, which can miss critical evidence distributed nonuniformly throughout the video. Moreover, they lack the ability to temporally localize such evidence in the broader context of the full video, which is required for answering complex questions. We propose a framework to enhance existing VideoQA datasets with evidence reasoning chains, automatically constructed by searching for optimal intervals of interest in the video with supporting evidence, that maximizes the likelihood of answering a given question. We train our model (VITED) to generate these evidence chains directly, enabling it to both localize evidence windows as well as perform multi-step reasoning across them in long-form video content. We show the value of our evidence-distilled models on a suite of long video QA benchmarks where we outperform state-of-the-art approaches that lack evidence reasoning capabilities.

Arjun Somayazulu, Efi Mavroudi, Changan Chen et al.

Traditional methods for view-invariant learning from video rely on controlled multi-view settings with minimal scene clutter. However, they struggle with in-the-wild videos that exhibit extreme viewpoint differences and share little visual content. We introduce a method for learning rich video representations in the presence of such severe view-occlusions. We first define a geometry-based metric that ranks views at a fine-grained temporal scale by their likely occlusion level. Then, using those rankings, we formulate a knowledge distillation objective that preserves action-centric semantics with a novel curriculum learning procedure that pairs incrementally more challenging views over time, thereby allowing smooth adaptation to extreme viewpoint differences. We evaluate our approach on two tasks, outperforming SOTA models on both temporal keystep grounding and fine-grained keystep recognition benchmarks - particularly on views that exhibit severe occlusion.

Shraman Pramanick, Effrosyni Mavroudi, Yale Song et al.

We introduce ED-VTG, a method for fine-grained video temporal grounding utilizing multi-modal large language models. Our approach harnesses the capabilities of multimodal LLMs to jointly process text and video, in order to effectively localize natural language queries in videos through a two-stage process. Rather than being directly grounded, language queries are initially transformed into enriched sentences that incorporate missing details and cues to aid in grounding. In the second stage, these enriched queries are grounded, using a lightweight decoder, which specializes at predicting accurate boundaries conditioned on contextualized representations of the enriched queries. To mitigate noise and reduce the impact of hallucinations, our model is trained with a multiple-instance-learning objective that dynamically selects the optimal version of the query for each training sample. We demonstrate state-of-the-art results across various benchmarks in temporal video grounding and paragraph grounding settings. Experiments reveal that our method significantly outperforms all previously proposed LLM-based temporal grounding approaches and is either superior or comparable to specialized models, while maintaining a clear advantage against them in zero-shot evaluation scenarios.

Jue Wang, Lorenzo Torresani

Video transformers have recently emerged as an effective alternative to convolutional networks for action classification. However, most prior video transformers adopt either global space-time attention or hand-defined strategies to compare patches within and across frames. These fixed attention schemes not only have high computational cost but, by comparing patches at predetermined locations, they neglect the motion dynamics in the video. In this paper, we introduce the Deformable Video Transformer (DVT), which dynamically predicts a small subset of video patches to attend for each query location based on motion information, thus allowing the model to decide where to look in the video based on correspondences across frames. Crucially, these motion-based correspondences are obtained at zero-cost from information stored in the compressed format of the video. Our deformable attention mechanism is optimised directly with respect to classification performance, thus eliminating the need for suboptimal hand-design of attention strategies. Experiments on four large-scale video benchmarks (Kinetics-400, Something-Something-V2, EPIC-KITCHENS and Diving-48) demonstrate that, compared to existing video transformers, our model achieves higher accuracy at the same or lower computational cost, and it attains state-of-the-art results on these four datasets.

Jerry Wei, Lorenzo Torresani, Jason Wei et al.

The classification of histopathology images fundamentally differs from traditional image classification tasks because histopathology images naturally exhibit a range of diagnostic features, resulting in a diverse range of annotator agreement levels. However, examples with high annotator disagreement are often either assigned the majority label or discarded entirely when training histopathology image classifiers. This widespread practice often yields classifiers that do not account for example difficulty and exhibit poor model calibration. In this paper, we ask: can we improve model calibration by endowing histopathology image classifiers with inductive biases about example difficulty? We propose several label smoothing methods that utilize per-image annotator agreement. Though our methods are simple, we find that they substantially improve model calibration, while maintaining (or even improving) accuracy. For colorectal polyp classification, a common yet challenging task in gastrointestinal pathology, we find that our proposed agreement-aware label smoothing methods reduce calibration error by almost 70%. Moreover, we find that using model confidence as a proxy for annotator agreement also improves calibration and accuracy, suggesting that datasets without multiple annotators can still benefit from our proposed label smoothing methods via our proposed confidence-aware label smoothing methods. Given the importance of calibration (especially in histopathology image analysis), the improvements from our proposed techniques merit further exploration and potential implementation in other histopathology image classification tasks.

Xudong Lin, Fabio Petroni, Gedas Bertasius et al.

In this paper we consider the problem of classifying fine-grained, multi-step activities (e.g., cooking different recipes, making disparate home improvements, creating various forms of arts and crafts) from long videos spanning up to several minutes. Accurately categorizing these activities requires not only recognizing the individual steps that compose the task but also capturing their temporal dependencies. This problem is dramatically different from traditional action classification, where models are typically optimized on videos that span only a few seconds and that are manually trimmed to contain simple atomic actions. While step annotations could enable the training of models to recognize the individual steps of procedural activities, existing large-scale datasets in this area do not include such segment labels due to the prohibitive cost of manually annotating temporal boundaries in long videos. To address this issue, we propose to automatically identify steps in instructional videos by leveraging the distant supervision of a textual knowledge base (wikiHow) that includes detailed descriptions of the steps needed for the execution of a wide variety of complex activities. Our method uses a language model to match noisy, automatically-transcribed speech from the video to step descriptions in the knowledge base. We demonstrate that video models trained to recognize these automatically-labeled steps (without manual supervision) yield a representation that achieves superior generalization performance on four downstream tasks: recognition of procedural activities, step classification, step forecasting and egocentric video classification.

Yiren Jian, Lorenzo Torresani

Few-shot classification requires adapting knowledge learned from a large annotated base dataset to recognize novel unseen classes, each represented by few labeled examples. In such a scenario, pretraining a network with high capacity on the large dataset and then finetuning it on the few examples causes severe overfitting. At the same time, training a simple linear classifier on top of "frozen" features learned from the large labeled dataset fails to adapt the model to the properties of the novel classes, effectively inducing underfitting. In this paper we propose an alternative approach to both of these two popular strategies. First, our method pseudo-labels the entire large dataset using the linear classifier trained on the novel classes. This effectively "hallucinates" the novel classes in the large dataset, despite the novel categories not being present in the base database (novel and base classes are disjoint). Then, it finetunes the entire model with a distillation loss on the pseudo-labeled base examples, in addition to the standard cross-entropy loss on the novel dataset. This step effectively trains the network to recognize contextual and appearance cues that are useful for the novel-category recognition but using the entire large-scale base dataset and thus overcoming the inherent data-scarcity problem of few-shot learning. Despite the simplicity of the approach, we show that that our method outperforms the state-of-the-art on four well-established few-shot classification benchmarks.

Kristen Grauman, Andrew Westbury, Eugene Byrne et al.

We introduce Ego4D, a massive-scale egocentric video dataset and benchmark suite. It offers 3,670 hours of daily-life activity video spanning hundreds of scenarios (household, outdoor, workplace, leisure, etc.) captured by 931 unique camera wearers from 74 worldwide locations and 9 different countries. The approach to collection is designed to uphold rigorous privacy and ethics standards with consenting participants and robust de-identification procedures where relevant. Ego4D dramatically expands the volume of diverse egocentric video footage publicly available to the research community. Portions of the video are accompanied by audio, 3D meshes of the environment, eye gaze, stereo, and/or synchronized videos from multiple egocentric cameras at the same event. Furthermore, we present a host of new benchmark challenges centered around understanding the first-person visual experience in the past (querying an episodic memory), present (analyzing hand-object manipulation, audio-visual conversation, and social interactions), and future (forecasting activities). By publicly sharing this massive annotated dataset and benchmark suite, we aim to push the frontier of first-person perception. Project page: https://ego4d-data.org/

Jue Wang, Gedas Bertasius, Du Tran et al.

Video transformers have recently emerged as a competitive alternative to 3D CNNs for video understanding. However, due to their large number of parameters and reduced inductive biases, these models require supervised pretraining on large-scale image datasets to achieve top performance. In this paper, we empirically demonstrate that self-supervised pretraining of video transformers on video-only datasets can lead to action recognition results that are on par or better than those obtained with supervised pretraining on large-scale image datasets, even massive ones such as ImageNet-21K. Since transformer-based models are effective at capturing dependencies over extended temporal spans, we propose a simple learning procedure that forces the model to match a long-term view to a short-term view of the same video. Our approach, named Long-Short Temporal Contrastive Learning (LSTCL), enables video transformers to learn an effective clip-level representation by predicting temporal context captured from a longer temporal extent. To demonstrate the generality of our findings, we implement and validate our approach under three different self-supervised contrastive learning frameworks (MoCo v3, BYOL, SimSiam) using two distinct video-transformer architectures, including an improved variant of the Swin Transformer augmented with space-time attention. We conduct a thorough ablation study and show that LSTCL achieves competitive performance on multiple video benchmarks and represents a convincing alternative to supervised image-based pretraining.

Xitong Yang, Haoqi Fan, Lorenzo Torresani et al.

The standard way of training video models entails sampling at each iteration a single clip from a video and optimizing the clip prediction with respect to the video-level label. We argue that a single clip may not have enough temporal coverage to exhibit the label to recognize, since video datasets are often weakly labeled with categorical information but without dense temporal annotations. Furthermore, optimizing the model over brief clips impedes its ability to learn long-term temporal dependencies. To overcome these limitations, we introduce a collaborative memory mechanism that encodes information across multiple sampled clips of a video at each training iteration. This enables the learning of long-range dependencies beyond a single clip. We explore different design choices for the collaborative memory to ease the optimization difficulties. Our proposed framework is end-to-end trainable and significantly improves the accuracy of video classification at a negligible computational overhead. Through extensive experiments, we demonstrate that our framework generalizes to different video architectures and tasks, outperforming the state of the art on both action recognition (e.g., Kinetics-400 & 700, Charades, Something-Something-V1) and action detection (e.g., AVA v2.1 & v2.2).

Leda Sarı, Kritika Singh, Jiatong Zhou et al.

Although speaker verification has conventionally been an audio-only task, some practical applications provide both audio and visual streams of input. In these cases, the visual stream provides complementary information and can often be leveraged in conjunction with the acoustics of speech to improve verification performance. In this study, we explore audio-visual approaches to speaker verification, starting with standard fusion techniques to learn joint audio-visual (AV) embeddings, and then propose a novel approach to handle cross-modal verification at test time. Specifically, we investigate unimodal and concatenation based AV fusion and report the lowest AV equal error rate (EER) of 0.7% on the VoxCeleb1 dataset using our best system. As these methods lack the ability to do cross-modal verification, we introduce a multi-view model which uses a shared classifier to map audio and video into the same space. This new approach achieves 28% EER on VoxCeleb1 in the challenging testing condition of cross-modal verification.

Jerry Wei, Arief Suriawinata, Bing Ren et al.

With the rise of deep learning, there has been increased interest in using neural networks for histopathology image analysis, a field that investigates the properties of biopsy or resected specimens traditionally manually examined under a microscope by pathologists. However, challenges such as limited data, costly annotation, and processing high-resolution and variable-size images make it difficult to quickly iterate over model designs. Throughout scientific history, many significant research directions have leveraged small-scale experimental setups as petri dishes to efficiently evaluate exploratory ideas. In this paper, we introduce a minimalist histopathology image analysis dataset (MHIST), an analogous petri dish for histopathology image analysis. MHIST is a binary classification dataset of 3,152 fixed-size images of colorectal polyps, each with a gold-standard label determined by the majority vote of seven board-certified gastrointestinal pathologists and annotator agreement level. MHIST occupies less than 400 MB of disk space, and a ResNet-18 baseline can be trained to convergence on MHIST in just 6 minutes using 3.5 GB of memory on a NVIDIA RTX 3090. As example use cases, we use MHIST to study natural questions such as how dataset size, network depth, transfer learning, and high-disagreement examples affect model performance. By introducing MHIST, we hope to not only help facilitate the work of current histopathology imaging researchers, but also make the field more-accessible to the general community. Our dataset is available at https://bmirds.github.io/MHIST.

Xudong Lin, Gedas Bertasius, Jue Wang et al.

We present \textsc{Vx2Text}, a framework for text generation from multimodal inputs consisting of video plus text, speech, or audio. In order to leverage transformer networks, which have been shown to be effective at modeling language, each modality is first converted into a set of language embeddings by a learnable tokenizer. This allows our approach to perform multimodal fusion in the language space, thus eliminating the need for ad-hoc cross-modal fusion modules. To address the non-differentiability of tokenization on continuous inputs (e.g., video or audio), we utilize a relaxation scheme that enables end-to-end training. Furthermore, unlike prior encoder-only models, our network includes an autoregressive decoder to generate open-ended text from the multimodal embeddings fused by the language encoder. This renders our approach fully generative and makes it directly applicable to different "video+$x$ to text" problems without the need to design specialized network heads for each task. The proposed framework is not only conceptually simple but also remarkably effective: experiments demonstrate that our approach based on a single architecture outperforms the state-of-the-art on three video-based text-generation tasks -- captioning, question answering and audio-visual scene-aware dialog.

Maxwell Mbabilla Aladago, Lorenzo Torresani

In contrast to traditional weight optimization in a continuous space, we demonstrate the existence of effective random networks whose weights are never updated. By selecting a weight among a fixed set of random values for each individual connection, our method uncovers combinations of random weights that match the performance of traditionally-trained networks of the same capacity. We refer to our networks as "slot machines" where each reel (connection) contains a fixed set of symbols (random values). Our backpropagation algorithm "spins" the reels to seek "winning" combinations, i.e., selections of random weight values that minimize the given loss. Quite surprisingly, we find that allocating just a few random values to each connection (e.g., 8 values per connection) yields highly competitive combinations despite being dramatically more constrained compared to traditionally learned weights. Moreover, finetuning these combinations often improves performance over the trained baselines. A randomly initialized VGG-19 with 8 values per connection contains a combination that achieves 91% test accuracy on CIFAR-10. Our method also achieves an impressive performance of 98.2% on MNIST for neural networks containing only random weights.

Joseph DiPalma, Arief A. Suriawinata, Laura J. Tafe et al.

Developing deep learning models to analyze histology images has been computationally challenging, as the massive size of the images causes excessive strain on all parts of the computing pipeline. This paper proposes a novel deep learning-based methodology for improving the computational efficiency of histology image classification. The proposed approach is robust when used with images that have reduced input resolution and can be trained effectively with limited labeled data. Pre-trained on the original high-resolution (HR) images, our method uses knowledge distillation (KD) to transfer learned knowledge from a teacher model to a student model trained on the same images at a much lower resolution. To address the lack of large-scale labeled histology image datasets, we perform KD in a self-supervised manner. We evaluate our approach on two histology image datasets associated with celiac disease (CD) and lung adenocarcinoma (LUAD). Our results show that a combination of KD and self-supervision allows the student model to approach, and in some cases, surpass the classification accuracy of the teacher, while being much more efficient. Additionally, we observe an increase in student classification performance as the size of the unlabeled dataset increases, indicating that there is potential to scale further. For the CD data, our model outperforms the HR teacher model, while needing 4 times fewer computations. For the LUAD data, our student model results at 1.25x magnification are within 3% of the teacher model at 10x magnification, with a 64 times computational cost reduction. Moreover, our CD outcomes benefit from performance scaling with the use of more unlabeled data. For 0.625x magnification, using unlabeled data improves accuracy by 4% over the baseline. Thus, our method can improve the feasibility of deep learning solutions for digital pathology with standard computational hardware.

Jerry Wei, Arief Suriawinata, Bing Ren et al.

Applying curriculum learning requires both a range of difficulty in data and a method for determining the difficulty of examples. In many tasks, however, satisfying these requirements can be a formidable challenge. In this paper, we contend that histopathology image classification is a compelling use case for curriculum learning. Based on the nature of histopathology images, a range of difficulty inherently exists among examples, and, since medical datasets are often labeled by multiple annotators, annotator agreement can be used as a natural proxy for the difficulty of a given example. Hence, we propose a simple curriculum learning method that trains on progressively-harder images as determined by annotator agreement. We evaluate our hypothesis on the challenging and clinically-important task of colorectal polyp classification. Whereas vanilla training achieves an AUC of 83.7% for this task, a model trained with our proposed curriculum learning approach achieves an AUC of 88.2%, an improvement of 4.5%. Our work aims to inspire researchers to think more creatively and rigorously when choosing contexts for applying curriculum learning.

Gedas Bertasius, Lorenzo Torresani

Many objects in the real world undergo dramatic variations in visual appearance. For example, a tomato may be red or green, sliced or chopped, fresh or fried, liquid or solid. Training a single detector to accurately recognize tomatoes in all these different states is challenging. On the other hand, contextual cues (e.g., the presence of a knife, a cutting board, a strainer or a pan) are often strongly indicative of how the object appears in the scene. Recognizing such contextual cues is useful not only to improve the accuracy of object detection or to determine the state of the object, but also to understand its functional properties and to infer ongoing or upcoming human-object interactions. A fully-supervised approach to recognizing object states and their contexts in the real-world is unfortunately marred by the long-tailed, open-ended distribution of the data, which would effectively require massive amounts of annotations to capture the appearance of objects in all their different forms. Instead of relying on manually-labeled data for this task, we propose a new framework for learning Contextualized OBject Embeddings (COBE) from automatically-transcribed narrations of instructional videos. We leverage the semantic and compositional structure of language by training a visual detector to predict a contextualized word embedding of the object and its associated narration. This enables the learning of an object representation where concepts relate according to a semantic language metric. Our experiments show that our detector learns to predict a rich variety of contextual object information, and that it is highly effective in the settings of few-shot and zero-shot learning.

Yongqin Xian, Bruno Korbar, Matthijs Douze et al.

Few-shot learning aims to recognize novel classes from a few examples. Although significant progress has been made in the image domain, few-shot video classification is relatively unexplored. We argue that previous methods underestimate the importance of video feature learning and propose to learn spatiotemporal features using a 3D CNN. Proposing a two-stage approach that learns video features on base classes followed by fine-tuning the classifiers on novel classes, we show that this simple baseline approach outperforms prior few-shot video classification methods by over 20 points on existing benchmarks. To circumvent the need of labeled examples, we present two novel approaches that yield further improvement. First, we leverage tag-labeled videos from a large dataset using tag retrieval followed by selecting the best clips with visual similarities. Second, we learn generative adversarial networks that generate video features of novel classes from their semantic embeddings. Moreover, we find existing benchmarks are limited because they only focus on 5 novel classes in each testing episode and introduce more realistic benchmarks by involving more novel classes, i.e. few-shot learning, as well as a mixture of novel and base classes, i.e. generalized few-shot learning. The experimental results show that our retrieval and feature generation approach significantly outperform the baseline approach on the new benchmarks.

Bruno Korbar, Fabio Petroni, Rohit Girdhar et al.

With the advent of large-scale multimodal video datasets, especially sequences with audio or transcribed speech, there has been a growing interest in self-supervised learning of video representations. Most prior work formulates the objective as a contrastive metric learning problem between the modalities. To enable effective learning, however, these strategies require a careful selection of positive and negative samples often combined with hand-designed curriculum policies. In this work we remove the need for negative sampling by taking a generative modeling approach that poses the objective as a translation problem between modalities. Such a formulation allows us to tackle a wide variety of downstream video understanding tasks by means of a single unified framework, without the need for large batches of negative samples common in contrastive metric learning. We experiment with the large-scale HowTo100M dataset for training, and report performance gains over the state-of-the-art on several downstream tasks including video classification (EPIC-Kitchens), question answering (TVQA), captioning (TVC, YouCook2, and MSR-VTT), and text-based clip retrieval (YouCook2 and MSR-VTT).

Jun Han, Fan Ding, Xianglong Liu et al.

Gradient-based approximate inference methods, such as Stein variational gradient descent (SVGD), provide simple and general-purpose inference engines for differentiable continuous distributions. However, existing forms of SVGD cannot be directly applied to discrete distributions. In this work, we fill this gap by proposing a simple yet general framework that transforms discrete distributions to equivalent piecewise continuous distributions, on which the gradient-free SVGD is applied to perform efficient approximate inference. The empirical results show that our method outperforms traditional algorithms such as Gibbs sampling and discontinuous Hamiltonian Monte Carlo on various challenging benchmarks of discrete graphical models. We demonstrate that our method provides a promising tool for learning ensembles of binarized neural network (BNN), outperforming other widely used ensemble methods on learning binarized AlexNet on CIFAR-10 dataset. In addition, such transform can be straightforwardly employed in gradient-free kernelized Stein discrepancy to perform goodness-of-fit (GOF) test on discrete distributions. Our proposed method outperforms existing GOF test methods for intractable discrete distributions.

Gedas Bertasius, Lorenzo Torresani

We introduce a method for simultaneously classifying, segmenting and tracking object instances in a video sequence. Our method, named MaskProp, adapts the popular Mask R-CNN to video by adding a mask propagation branch that propagates frame-level object instance masks from each video frame to all the other frames in a video clip. This allows our system to predict clip-level instance tracks with respect to the object instances segmented in the middle frame of the clip. Clip-level instance tracks generated densely for each frame in the sequence are finally aggregated to produce video-level object instance segmentation and classification. Our experiments demonstrate that our clip-level instance segmentation makes our approach robust to motion blur and object occlusions in video. MaskProp achieves the best reported accuracy on the YouTube-VIS dataset, outperforming the ICCV 2019 video instance segmentation challenge winner despite being much simpler and using orders of magnitude less labeled data (1.3M vs 1B images and 860K vs 14M bounding boxes).

Ruohan Gao, Tae-Hyun Oh, Kristen Grauman et al.

In the face of the video data deluge, today's expensive clip-level classifiers are increasingly impractical. We propose a framework for efficient action recognition in untrimmed video that uses audio as a preview mechanism to eliminate both short-term and long-term visual redundancies. First, we devise an ImgAud2Vid framework that hallucinates clip-level features by distilling from lighter modalities---a single frame and its accompanying audio---reducing short-term temporal redundancy for efficient clip-level recognition. Second, building on ImgAud2Vid, we further propose ImgAud-Skimming, an attention-based long short-term memory network that iteratively selects useful moments in untrimmed videos, reducing long-term temporal redundancy for efficient video-level recognition. Extensive experiments on four action recognition datasets demonstrate that our method achieves the state-of-the-art in terms of both recognition accuracy and speed.

Humam Alwassel, Dhruv Mahajan, Bruno Korbar et al.

Visual and audio modalities are highly correlated, yet they contain different information. Their strong correlation makes it possible to predict the semantics of one from the other with good accuracy. Their intrinsic differences make cross-modal prediction a potentially more rewarding pretext task for self-supervised learning of video and audio representations compared to within-modality learning. Based on this intuition, we propose Cross-Modal Deep Clustering (XDC), a novel self-supervised method that leverages unsupervised clustering in one modality (e.g., audio) as a supervisory signal for the other modality (e.g., video). This cross-modal supervision helps XDC utilize the semantic correlation and the differences between the two modalities. Our experiments show that XDC outperforms single-modality clustering and other multi-modal variants. XDC achieves state-of-the-art accuracy among self-supervised methods on multiple video and audio benchmarks. Most importantly, our video model pretrained on large-scale unlabeled data significantly outperforms the same model pretrained with full-supervision on ImageNet and Kinetics for action recognition on HMDB51 and UCF101. To the best of our knowledge, XDC is the first self-supervised learning method that outperforms large-scale fully-supervised pretraining for action recognition on the same architecture.

Laura Sevilla-Lara, Shengxin Zha, Zhicheng Yan et al.

Understanding temporal information and how the visual world changes over time is a fundamental ability of intelligent systems. In video understanding, temporal information is at the core of many current challenges, including compression, efficient inference, motion estimation or summarization. However, in current video datasets it has been observed that action classes can often be recognized without any temporal information from a single frame of video. As a result, both benchmarking and training in these datasets may give an unintentional advantage to models with strong image understanding capabilities, as opposed to those with strong temporal understanding. In this paper we address this problem head on by identifying action classes where temporal information is actually necessary to recognize them and call these "temporal classes". Selecting temporal classes using a computational method would bias the process. Instead, we propose a methodology based on a simple and effective human annotation experiment. We remove just the temporal information by shuffling frames in time and measure if the action can still be recognized. Classes that cannot be recognized when frames are not in order are included in the temporal Dataset. We observe that this set is statistically different from other static classes, and that performance in it correlates with a network's ability to capture temporal information. Thus we use it as a benchmark on current popular networks, which reveals a series of interesting facts. We also explore the effect of training on the temporal dataset, and observe that this leads to better generalization in unseen classes, demonstrating the need for more temporal data. We hope that the proposed dataset of temporal categories will help guide future research in temporal modeling for better video understanding.

Yufei Wang, Du Tran, Lorenzo Torresani

Although a video is effectively a sequence of images, visual perception systems typically model images and videos separately, thus failing to exploit the correlation and the synergy provided by these two media. While a few prior research efforts have explored the benefits of leveraging still-image datasets for video analysis, or vice-versa, most of these attempts have been limited to pretraining a model on one type of visual modality and then adapting it via finetuning on the other modality. In contrast, in this paper we introduce a framework that enables joint training of a unified model on mixed collections of image and video examples spanning different tasks. The key ingredient in our architecture design is a new network block, which we name UniDual. It consists of a shared 2D spatial convolution followed by two parallel point-wise convolutional layers, one devoted to images and the other one used for videos. For video input, the point-wise filtering implements a temporal convolution. For image input, it performs a pixel-wise nonlinear transformation. Repeated stacking of such blocks gives rise to a network where images and videos undergo partially distinct execution pathways, unified by spatial convolutions (capturing commonalities in visual appearance) but separated by point-wise operations (modeling patterns specific to each modality). Extensive experiments on Kinetics and ImageNet demonstrate that our UniDual model jointly trained on these datasets yields substantial accuracy gains for both tasks, compared to 1) training separate models, 2) traditional multi-task learning and 3) the conventional framework of pretraining-followed-by-finetuning. On Kinetics, the UniDual architecture applied to a state-of-the-art video backbone model (R(2+1)D-152) yields an additional video@1 accuracy gain of 1.5%.

Heng Wang, Du Tran, Lorenzo Torresani et al.

Motion is a salient cue to recognize actions in video. Modern action recognition models leverage motion information either explicitly by using optical flow as input or implicitly by means of 3D convolutional filters that simultaneously capture appearance and motion information. This paper proposes an alternative approach based on a learnable correlation operator that can be used to establish frame-toframe matches over convolutional feature maps in the different layers of the network. The proposed architecture enables the fusion of this explicit temporal matching information with traditional appearance cues captured by 2D convolution. Our correlation network compares favorably with widely-used 3D CNNs for video modeling, and achieves competitive results over the prominent two-stream network while being much faster to train. We empirically demonstrate that correlation networks produce strong results on a variety of video datasets, and outperform the state of the art on four popular benchmarks for action recognition: Kinetics, Something-Something, Diving48 and Sports1M.

Yang Wang, Vinh Tran, Gedas Bertasius et al.

We propose a method for human action recognition, one that can localize the spatiotemporal regions that `define' the actions. This is a challenging task due to the subtlety of human actions in video and the co-occurrence of contextual elements. To address this challenge, we utilize conjugate samples of human actions, which are video clips that are contextually similar to human action samples but do not contain the action. We introduce a novel attentional mechanism that can spatially and temporally separate human actions from the co-occurring contextual factors. The separation of the action and context factors is weakly supervised, eliminating the need for laboriously detailed annotation of these two factors in training samples. Our method can be used to build human action classifiers with higher accuracy and better interpretability. Experiments on several human action recognition datasets demonstrate the quantitative and qualitative benefits of our approach.