Recent advances in foundational 3D reconstruction models, such as DUSt3R and MASt3R, have shown great potential in 2D and 3D correspondence in static scenes. In this paper, we propose to adapt them for the task of point tracking through 3D grounded correspondence. We first demonstrate that these models are competitive point trackers when focusing on static points, present in current point tracking benchmarks (+33.5% on EgoPoints vs. CoTracker2). We propose to combine the reconstruction loss with training for dynamic correspondence along with a visibility head, and fine-tuning MASt3R for point tracking using a relatively small amount of synthetic data. Importantly, we only train and evaluate on pairs of frames where one contains the query point, effectively removing any temporal context.Using a mix of dynamic and static point correspondences, we achieve competitive or superior point tracking results on four datasets (e.g. competitive on TAP-Vid-DAVIS 73.8 δavg/ 85.8% occlusion acc. for PointSt3R compared to 75.7 / 88.3% for CoTracker2; and significantly outperform CoTracker3 on EgoPoints 61.3 vs 54.2 and RGB-S 87.0 vs 82.8). We also present results on 3D point tracking along with several ablations on training datasets and percentage of dynamic correspondences.

As humans move around, performing their daily tasks, they are able to recall where they have positioned objects in their environment, even if these objects are currently out of their sight. In this paper, we aim to mimic this spatial cognition ability. We thus formulate the task of Out of Sight, Not Out of Mind - 3D tracking active objects using observations captured through an egocentric camera. We introduce a simple but effective approach to address this challenging problem, called Lift, Match, and Keep (LMK). LMK lifts partial 2D observations to 3D world coordinates, matches them over time using visual appearance, 3D location and interactions to form object tracks, and keeps these object tracks even when they go out-of-view of the camera. We benchmark LMK on 100 long videos from EPICKITCHENS. Our results demonstrate that spatial cognition is critical for correctly locating objects over short and long time scales. E.g., for one long egocentric video, we estimate the 3D location of 50 active objects. After 120 seconds, 57 % of the objects are correctly localised by LMK, compared to just 33% by a recent 3D method for egocentric videos and 17 % by a general 2D tracking method.

We address the challenge of representation learning from a continuous stream of video as input, in a self-supervised manner. This differs from the standard approaches to video learning where videos are chopped and shuffled during training in order to create a non-redundant batch that satisfies the independently and identically distributed (IID) sample assumption expected by conventional training paradigms. When videos are only available as a continuous stream of input, the IID assumption is evidently broken, leading to poor performance. We demonstrate the drop in performance when moving from shuffled to sequential learning on three tasks: the one-video representation learning method DoRA, standard VideoMAE on multi-video datasets, and the task of future video prediction.To address this drop, we propose a geometric modification to standard optimizers, to decorrelate batches by utilising orthogonal gradients during training. The proposed modification can be applied to any optimizer –we demonstrate it with Stochastic Gradient Descent (SGD) and AdamW. Our proposed orthogonal optimizer allows models trained from streaming videos to alleviate the drop in representation learning performance, as evaluated on downstream tasks. On three scenarios (DoRA, VideoMAE, future prediction), we show our orthogonal optimizer outperforms the strong AdamW in all three scenarios.

We present a validation dataset of newly-collected kitchen-based egocentric videos, manually annotated with highly detailed and interconnected ground-truth labels covering: recipe steps, fine-grained actions, ingredients with nutritional values, moving objects, and audio annotations. Importantly, all annotations are grounded in 3D through digital twinning of the scene, fixtures, object locations, and primed with gaze. Footage is collected from unscripted recordings in diverse home environments, making HD-EPIC the first dataset collected in-the-wild but with detailed annotations matching those in controlled lab environments.We show the potential of our highly-detailed annotations through a challenging VQA benchmark of 26K questions assessing the capability to recognise recipes, ingredients, nutrition, fine-grained actions, 3D perception, object motion, and gaze direction. The powerful long-context Gemini Pro only achieves 37.6% on this benchmark, showcasing its difficulty and highlighting shortcomings in current VLMs. We additionally assess action recognition, sound recognition, and long-term video-object segmentation on HD-EPIC.HD-EPIC is 41 hours of video in 9 kitchens with digital twins of 413 kitchen fixtures, capturing 69 recipes, 59K fine-grained actions, 51K audio events, 20K object movements and 37K object masks lifted to 3D. On average, we have 263 annotations per minute of our unscripted videos.

Large-scale multimodal representation learning successfully optimizes for zero-shot transfer at test time. Yet the standard pretraining paradigm (contrastive learning on large amounts of image-text data) does not explicitly encourage representations to support few-shot adaptation. In this work, we propose a simple, but carefully designed extension to multimodal pretraining which enables representations to accommodate additional context. Using this objective, we show that vision-language models can be trained to exhibit significantly increased few-shot adaptation: across 21 downstream tasks, we find up to fourfold improvements in test-time sample efficiency, and average few-shot adaptation gains of over 5%, while retaining zero-shot generalization performance across model scales and training durations. In particular, equipped with simple, training-free, metric-based adaptation mechanisms, our representations easily surpass more complex and expensive optimization-based schemes, vastly simplifying generalization to new domains.

The goal of this work is to generate step-by-step visual instructions in the form of a sequence of images, given an input image that provides the scene context and the sequence of textual instructions. This is a challenging problem as it requires generating multi-step image sequences to achieve a complex goal while being grounded in a specific environment. Part of the challenge stems from the lack of large-scale training data for this problem. The contribution of this work is thus three-fold. First, we introduce an automatic approach for collecting large step-by-step visual instruction training data from instructional videos. We apply this approach to one million videos and create a large-scale, high-quality dataset of 0.6M sequences of image-text pairs. Second, we develop and train ShowHowTo, a video diffusion model capable of generating step-by-step visual instructions consistent with the provided input image. Third, we evaluate the generated image sequences across three dimensions of accuracy (step, scene, and task) and show our model achieves state-of-the-art results on all of them. Our code, dataset, and trained models are publicly available.

We introduce EgoPoints, a benchmark for point tracking in egocentric videos. We annotate 4.7K challenging tracks in egocentric sequences. Compared to the popular TAP-Vid-DAVIS [7] evaluation benchmark, we include 9x more points that go out-of-view and 59x more points that require re-identification (ReID) after returning to view. To measure the performance of models on these challenging points, we introduce evaluation metrics that specifically monitor tracking performance on points in-view, out-of-view, and points that require re-identification. We then propose a pipeline to create semi-real sequences, with automatic ground truth. We generate 11 K such sequences by combining dynamic Kubric [17] objects with scene points from EPIC Fields [38]. When fine-tuning point tracking methods on these sequences and evaluating on our annotated EgoPoints sequences, we improve Co-Tracker [22] across all metrics, including the tracking accuracy$\delta^{\ast}_{avg}$by 2.7 percentage points and accuracy on ReID sequences ($ReID\delta_{avg}$) by 2.4 points. We also improve$\delta^{\ast}_{avg}$and$ReID\delta_{avg}$of PIPs++ [42] by 0.3 and 2.8 respectively.

Video Moment Retrieval is a common task to evaluate the performance of visual-language models-it involves localising start and end times of moments in videos from query sentences. The current task formulation assumes that the queried moment is present in the video, resulting in false positive moment predictions when irrelevant query sentences are provided. In this paper we propose the task of Negative-Aware Video Moment Retrieval (NA-VMR), which considers both moment retrieval accuracy and negative query rejection accuracy. We make the distinction between In-Domain and Out-of-Domain negative queries and provide new evaluation benchmarks for two popular video moment retrieval datasets: QVHighlights and Charades-STA. We analyse the ability of current SOTA video moment retrieval approaches to adapt to Negative-Aware Video Moment Retrieval and propose UniVTG-NA, an adaptation of UniVTG designed to tackle NA-VMR. UniVTG-NA achieves high negative rejection accuracy (avg. 98.4%) scores while retaining moment retrieval scores to within 3.87% Recall@1. Dataset splits are available at https://github.com/keflanagan/MomentofUntruth.

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. https://ego-exo4d-data.org/

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.

We introduce EPIC-SOUNDS, a large-scale dataset of audio annotations capturing temporal extents and class labels within the audio stream of the egocentric videos. We propose an annotation pipeline where annotators temporally label distinguishable audio segments and describe the action that could have caused this sound. We identify actions that can be discriminated purely from audio, through grouping these free-form descriptions of audio into classes. For actions that involve objects colliding, we collect human annotations of the materials of these objects (e.g., a glass object being placed on a wooden surface), which we verify from video, discarding ambiguities. Overall, EPIC-SOUNDS includes 78.4 k categorised segments of audible events and actions, distributed across 44 classes as well as 39.2 k non-categorised segments. We train and evaluate state-of-the-art audio recognition and detection models on our dataset, for both audio-only and audio-visual methods. We also conduct analysis on: the temporal overlap between audio events, the temporal and label correlations between audio and visual modalities, the ambiguities in annotating materials from audio-only input, the importance of audio-only labels and the limitations of current models to understand actions that sound.

We introduce a framework for online learning from a single continuous video stream-the way people and animals learn, without mini-batches, data augmentation or shuffling. This poses great challenges given the high correlation between consecutive video frames and there is very little prior work on it. Our framework allows us to do a first deep dive into the topic and includes a collection of streams and tasks composed from two existing video datasets, plus methodology for performance evaluation that considers both adaptation and generalization. We employ pixel-to-pixel modelling as a practical and flexible way to switch between pre-training and single-stream evaluation as well as between arbitrary tasks, without ever requiring changes to models and always using the same pixel loss. Equipped with this framework we obtained large single-stream learning gains from pre-training with a novel family of future prediction tasks, found that momentum hurts, and that the pace of weight updates matters. The combination of these insights leads to matching the performance of IID learning with batch size 1, when using the same architecture and without costly replay buffers. An overview of the pa-per is available online at https://sites.google.com/view/one-stream-video.

Diverse actions give rise to rich audio-visual signals in long videos. Recent works showcase that the two modalities of audio and video exhibit different temporal extents of events and distinct labels. We address the interplay between the two modalities in long videos by explicitly modelling the temporal extents of audio and visual events. We propose the Time Interval Machine (TIM) where a modality-specific time interval poses as a query to a transformer encoder that ingests a long video input. The encoder then attends to the specified interval, as well as the surrounding context in both modalities, in order to recognise the ongoing action. We test TIM on three long audio-visual video datasets: EPIC-KITCHENS, Perception Test, and AVE, reporting state-of-the-art (SOTA) for recognition. On EPIC-KITCHENS, we beat previous SOTA that utilises LLMs and significantly larger pre-training by 2.9% top-1 action recognition accuracy. Additionally, we show that TIM can be adapted for action detection, using dense multi-scale inter-val queries, outperforming SOTA on EPIC-KITCHENS-IOO for most metrics, and showing strong performance on the Perception Test. Our ablations show the critical role of in-tegrating the two modalities and modelling their time inter-vals in achieving this performance. Code and models at: https://github.com/JacobChalk/TIM.

We present Ego-Exo4D, a diverse, large-scale multi-modal multiview video dataset and benchmark challenge. Ego-Exo4D centers around simultaneously-captured ego-centric 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 un-precedented: 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.

We address the task of generating temporally consistent and physically plausible images of actions and object state transformations. Given an input image and a text prompt describing the targeted transformation, our generated images preserve the environment and transform objects in the initial image. Our contributions are threefold. First, we leverage a large body of instructional videos and automatically mine a dataset of triplets of consecutive frames corresponding to initial object states, actions, and resulting object transformations. Second, equipped with this data, we develop and train a conditioned diffusion model dubbed GenHowTo. Third, we evaluate GenHowTo on a variety of objects and actions and show superior performance compared to existing methods. In particular, we introduce a quantitative evaluation where GenHowTo achieves 88% and 74% on seen and unseen interaction categories, respectively, outperforming prior work by a large margin.

Teaching robots novel skills with demonstrations via human-in-the-loop data collection techniques like kinesthetic teaching or teleoperation puts a heavy burden on human supervisors. In contrast to this paradigm, it is often significantly easier to provide raw, action-free visual data of tasks being performed. Moreover, this data can even be mined from video datasets or the web. Ideally, this data can serve to guide robot learning for new tasks in novel environments, informing both "what" to do and "how" to do it. A powerful way to encode both the "what" and the "how" is to infer a well-shaped reward function for reinforcement learning. The challenge is determining how to ground visual demonstration inputs into a well-shaped and informative reward function. We propose a technique Rank2Reward for learning behaviors from videos of tasks being performed without access to any low-level states and actions. We do so by leveraging the videos to learn a reward function that measures incremental "progress" through a task by learning how to temporally rank the video frames in a demonstration. By inferring an appropriate ranking, the reward function is able to guide reinforcement learning by indicating when task progress is being made. This ranking function can be integrated into an adversarial imitation learning scheme resulting in an algorithm that can learn behaviors without exploiting the learned reward function. We demonstrate the effectiveness of Rank2Reward at learning behaviors from raw video on a number of tabletop manipulation tasks in both simulations and on a real-world robotic arm. We also demonstrate how Rank2Reward can be easily extended to be applicable to web-scale video datasets. Code and videos are available at https://rank2reward.github.io

Abstract What will the future be? We wonder! In this survey, we explore the gap between current research in egocentric vision and the ever-anticipated future, where wearable computing, with outward facing cameras and digital overlays, is expected to be integrated in our every day lives. To understand this gap, the article starts by envisaging the future through character-based stories, showcasing through examples the limitations of current technology. We then provide a mapping between this future and previously defined research tasks. For each task, we survey its seminal works, current state-of-the-art methodologies and available datasets, then reflect on shortcomings that limit its applicability to future research. Note that this survey focuses on software models for egocentric vision, independent of any specific hardware. The paper concludes with recommendations for areas of immediate explorations so as to unlock our path to the future always-on, personalised and life-enhancing egocentric vision.

This paper presents an investigation into long-tail video recognition. We demonstrate that, unlike naturally-collected video datasets and existing long-tail image benchmarks, current video benchmarks fall short on multiple long-tailed properties. Most critically, they lack few-shot classes in their tails. In response, we propose new video benchmarks that better assess long-tail recognition, by sampling subsets from two datasets: SSv2 and VideoLT. We then propose a method, Long-Tail Mixed Reconstruction (LMR), which reduces overfitting to instances from few-shot classes by reconstructing them as weighted combinations of samples from head classes. LMR then employs label mixing to learn robust decision boundaries. It achieves state-of-the-art average class accuracy on EPIC-KITCHENS and the proposed SSv2-LT and VideoLT-LT. Benchmarks and code at: github.com/tobyperrett/lmr

Early action prediction deals with inferring the ongoing action from partially-observed videos, typically at the outset of the video. We propose a bottleneck-based attention model that captures the evolution of the action, through progressive sampling over fine-to-coarse scales. Our proposed Temporal Progressive (TemPr) model is composed of multiple attention towers, one for each scale. The predicted action label is based on the collective agreement considering confidences of these towers. Extensive experiments over four video datasets showcase state-of-the-art performance on the task of Early Action Prediction across a range of encoder architectures. We demonstrate the effectiveness and consistency of TemPr through detailed ablations.††Code is available at: https://tinyurl.com/temprog

We introduce EPIC-SOUNDS, a large-scale dataset of audio annotations capturing temporal extents and class labels within the audio stream of the egocentric videos from EPIC-KITCHENS-100. We propose an annotation pipeline where annotators temporally label distinguishable audio segments and describe the action that could have caused this sound. We identify actions that can be discriminated purely from audio, through grouping free-form descriptions into classes. For actions that involve objects colliding, we collect human annotations of the materials of these objects (e.g. a glass object being placed on a wooden surface), which we verify from visual labels, discarding ambiguities. Overall, EPIC-SOUNDS includes 78.4k categorised segments of audible events and actions, distributed across 44 classes, as well as 39.2k non-categorised segments, totalling 117.6k segments spanning 100 hours of audio, capturing diverse actions that sound in home kitchens. We train and evaluate two state-of-the-art audio recognition models on our dataset, highlighting the importance of audio-only labels and the limitations of current models to recognise actions that sound.EPIC-SOUNDS and baseline source code is available from: https://epic-kitchens.github.io/epic-sounds.

A key function of auditory cognition is the association of characteristic sounds with their corresponding semantics over time. Humans attempting to discriminate between fine-grained audio categories, often replay the same discriminative sounds to increase their prediction confidence. We propose an end-to-end attention-based architecture that through selective repetition attends over the most discriminative sounds across the audio sequence. Our model initially uses the full audio sequence and iteratively refines the temporal segments replayed based on slot attention. At each playback, the selected segments are replayed using a smaller hop length which represents higher resolution features within these segments. We show that our method can consistently achieve state-of-the-art performance across three audio-classification benchmarks: AudioSet, VGG-Sound, and EPIC-KITCHENS-100.1

We propose and address a new generalisation problem: can a model trained for action recognition successfully classify actions when they are performed within a previously unseen scenario and in a previously unseen location? To answer this question, we introduce the Action Recognition Generalisation Over scenarios and locations dataset (ARGO1M), which contains 1.1M video clips from the large-scale Ego4D dataset, across 10 scenarios and 13 locations. We demonstrate recognition models struggle to generalise over 10 proposed test splits, each of an unseen scenario in an unseen location. We thus propose CIR, a method to represent each video as a Cross-Instance Reconstruction of videos from other domains. Reconstructions are paired with text narrations to guide the learning of a domain generalisable representation. We provide extensive analysis and ablations on ARGO1M that show CIR outperforms prior domain generalisation works on all test splits. Code and data: https://chiaraplizz.github.io/what-can-a-cook/.

We propose a novel multimodal video benchmark - the Perception Test - to evaluate the perception and reasoning skills of pre-trained multimodal models (e.g. Flamingo, BEiT-3, or GPT-4). Compared to existing benchmarks that focus on computational tasks (e.g. classification, detection or tracking), the Perception Test focuses on skills (Memory, Abstraction, Physics, Semantics) and types of reasoning (descriptive, explanatory, predictive, counterfactual) across video, audio, and text modalities, to provide a comprehensive and efficient evaluation tool. The benchmark probes pre-trained models for their transfer capabilities, in a zero-shot / few-shot or limited finetuning regime. For these purposes, the Perception Test introduces 11.6k real-world videos, 23s average length, designed to show perceptually interesting situations, filmed by around 100 participants worldwide. The videos are densely annotated with six types of labels (multiple-choice and grounded video question-answers, object and point tracks, temporal action and sound segments), enabling both language and non-language evaluations. The fine-tuning and validation splits of the benchmark are publicly available (CC-BY license), in addition to a challenge server with a held-out test split. Human baseline results compared to state-of-the-art video QA models show a significant gap in performance (91.4% vs 45.8%), suggesting that there is significant room for improvement in multimodal video understanding.Dataset, baselines code, and challenge server are available at https://github.com/deepmind/perception_test

Neural rendering is fuelling a unification of learning, 3D geometry and video understanding that has been waiting for more than two decades. Progress, however, is still hampered by a lack of suitable datasets and benchmarks. To address this gap, we introduce EPIC Fields, an augmentation of EPIC-KITCHENS with 3D camera information. Like other datasets for neural rendering, EPIC Fields removes the complex and expensive step of reconstructing cameras using photogrammetry, and allows researchers to focus on modelling problems. We illustrate the challenge of photogrammetry in egocentric videos of dynamic actions and propose innovations to address them. Compared to other neural rendering datasets, EPIC Fields is better tailored to video understanding because it is paired with labelled action segments and the recent VISOR segment annotations. To further motivate the community, we also evaluate two benchmark tasks in neural rendering and segmenting dynamic objects, with strong baselines that showcase what is not possible today. We also highlight the advantage of geometry in semi-supervised video object segmentations on the VISOR annotations. EPIC Fields reconstructs 96\% of videos in EPIC-KITCHENS, registering 19M frames in 99 hours recorded in 45 kitchens, and is available from: http://epic-kitchens.github.io/epic-fields

The papers in this special issue focus on egocentric perception. It gathers recent advances in this field that brings together multiple communities including computer vision, machine learning, and multimedia.

This paper proposes an interaction reasoning network for modelling spatio-temporal relationships between hands and objects in egocentric video. The proposed interaction unit utilises a Transformer-style module to reason about each acting hand, and its spatio-temporal relations to the other hand as well as objects being interacted with. We show that modelling two-handed interactions are critical for action recognition in egocentric video, and demonstrate that by using positionally-encoded trajectories, the network can better recognise observed interactions. We train and evaluate our proposed network on large-scale egocentric EPIC-KITCHENS-100 and crowd-sourced Something-Else datasets, with an ablation study to showcase our proposal.

Current one-stage action detection methods, which simultaneously predict action boundaries and the corresponding class, do not estimate or use a measure of confidence in their boundary predictions, which can lead to inaccurate boundaries. We incorporate the estimation of boundary confidence into one-stage anchor-free detection, through an additional prediction head that predicts the refined boundaries with higher confidence. We obtain state-of-the-art performance on the challenging EPICKITCHENS-100 action detection as well as the standard THUMOS14 action detection benchmarks, and achieve improvement on the ActivityNet-1.3 benchmark.

We introduce Ego4D, a massive-scale egocentric video dataset and benchmark suite. It offers 3,670 hours of dailylife 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/

Our lives can be seen as a complex weaving of activities; we switch from one activity to another, to maximise our achievements or in reaction to demands placed upon us. Observing a video of unscripted daily activities, we parse the video into its constituent activity threads through a process we call unweaving. To accomplish this, we introduce a video representation explicitly capturing activity threads called a thread bank, along with a neural controller capable of detecting goal changes and resuming of past activities, together forming UnweaveNet. We train and evaluate UnweaveNet on sequences from the unscripted egocentric dataset EPIC-KITCHENS. We propose and showcase the efficacy of pretraining UnweaveNet in a self-supervised manner.

We introduce VISOR, a new dataset of pixel annotations and a benchmark suite for segmenting hands and active objects in egocentric video. VISOR annotates videos from EPIC-KITCHENS, which comes with a new set of challenges not encountered in current video segmentation datasets. Specifically, we need to ensure both short- and long-term consistency of pixel-level annotations as objects undergo transformative interactions, e.g. an onion is peeled, diced and cooked - where we aim to obtain accurate pixel-level annotations of the peel, onion pieces, chopping board, knife, pan, as well as the acting hands. VISOR introduces an annotation pipeline, AI-powered in parts, for scalability and quality. In total, we publicly release 272K manual semantic masks of 257 object classes, 9.9M interpolated dense masks, 67K hand-object relations, covering 36 hours of 179 untrimmed videos. Along with the annotations, we introduce three challenges in video object segmentation, interaction understanding and long-term reasoning.For data, code and leaderboards: http://epic-kitchens.github.io/VISOR

Video-Language Pretraining (VLP), which aims to learn transferable representation to advance a wide range of video-text downstream tasks, has recently received increasing attention. Best performing works rely on large-scale, 3rd-person video-text datasets, such as HowTo100M. In this work, we exploit the recently released Ego4D dataset to pioneer Egocentric VLP along three directions. (i) We create EgoClip, a 1st-person video-text pretraining dataset comprising 3.8M clip-text pairs well-chosen from Ego4D, covering a large variety of human daily activities. (ii) We propose a novel pretraining objective, dubbed EgoNCE, which adapts video-text contrastive learning to the egocentric domain by mining egocentric-aware positive and negative samples. (iii) We introduce EgoMCQ, a development benchmark that is close to EgoClip and hence can support effective validation and fast exploration of our design decisions in EgoClip and EgoNCE. Furthermore, we demonstrate strong performance on five egocentric downstream tasks across three datasets: video-text retrieval on EPIC-KITCHENS-100; action recognition on Charades-Ego; natural language query, moment query, and object state change classification on Ego4D challenge benchmarks. The dataset and code are available at https://github.com/showlab/EgoVLP.

Existing video summarisation techniques are quite generic in nature, since they generally overlook the important aspect of what actual purpose the summary will be serving. In sharp contrast with this mainstream work, it can be acknowledged that there are many possible purposes the same videos can be summarised for. Accordingly, we consider a novel perspective: summaries with a purpose. This work is an attempt to both, call the attention on this neglected aspect of video summarisation research, and to illustrate it and explore it with two concrete purposes, focusing on first-person-view videos. The proposed purpose-oriented summarisation techniques are framed under the common (frame-level) scoring and selection paradigm, and have been tested on two egocentric datasets, BEOID and EGTEA-Gaze+. The necessary purpose-specific evaluation metrics are also introduced. The proposed approach is compared with two purpose-agnostic summarisation baselines. On the one hand, a partially agnostic method uses the scores obtained by the proposed approach, but follows a standard generic frame selection technique. On the other hand, the fully agnostic method do not use any purpose-based information, and relies on generic concepts such as diversity and representativeness. The results of the experimental work show that the proposed approaches compare favourably with respect to both baselines. More specifically, the purpose-specific approach generally produces summaries with the best compromise between summary lengths and favourable purpose-specific metrics. Interestingly, it is also observed that results of the partially-agnostic baseline tend to be better than those of the fully-agnostic one. These observations provide strong evidence on the advantage and relevance of purpose-specific summarisation techniques and evaluation metrics, and encourage further work on this important subject.

Abstract This paper introduces the pipeline to extend the largest dataset in egocentric vision, EPIC-KITCHENS. The effort culminates in EPIC-KITCHENS-100, a collection of 100 hours, 20M frames, 90K actions in 700 variable-length videos, capturing long-term unscripted activities in 45 environments, using head-mounted cameras. Compared to its previous version (Damen in Scaling egocentric vision: ECCV, 2018), EPIC-KITCHENS-100 has been annotated using a novel pipeline that allows denser (54% more actions per minute) and more complete annotations of fine-grained actions (+128% more action segments). This collection enables new challenges such as action detection and evaluating the “test of time”—i.e. whether models trained on data collected in 2018 can generalise to new footage collected two years later. The dataset is aligned with 6 challenges: action recognition (full and weak supervision), action detection, action anticipation, cross-modal retrieval (from captions), as well as unsupervised domain adaptation for action recognition. For each challenge, we define the task, provide baselines and evaluation metrics.

We propose a novel approach to few-shot action recognition, finding temporally-corresponding frame tuples between the query and videos in the support set. Distinct from previous few-shot works, we construct class prototypes using the CrossTransformer attention mechanism to observe relevant sub-sequences of all support videos, rather than using class averages or single best matches. Video representations are formed from ordered tuples of varying numbers of frames, which allows sub-sequences of actions at different speeds and temporal offsets to be compared.1Our proposed Temporal-Relational CrossTransformers (TRX) achieve state-of-the-art results on few-shot splits of Kinetics, Something-Something V2 (SSv2), HMDB51 and UCF101. Importantly, our method outperforms prior work on SSv2 by a wide margin (12%) due to the its ability to model temporal relations. A detailed ablation showcases the importance of matching to multiple support set videos and learning higher-order relational CrossTransformers.

Current video retrieval efforts all found their evaluation on an instance-based assumption, that only a single caption is relevant to a query video and vice versa. We demonstrate that this assumption results in performance comparisons often not indicative of models’ retrieval capabilities. We propose a move to semantic similarity video retrieval, where (i) multiple videos/captions can be deemed equally relevant, and their relative ranking does not affect a method’s reported performance and (ii) retrieved videos/captions are ranked by their similarity to a query. We propose several proxies to estimate semantic similarities in large-scale retrieval datasets, without additional annotations. Our analysis is performed on three commonly used video retrieval datasets (MSR-VTT, YouCook2 and EPIC-KITCHENS).

We propose a two-stream convolutional network for audio recognition, that operates on time-frequency spectrogram inputs. Following similar success in visual recognition, we learn Slow-Fast auditory streams with separable convolutions and multi-level lateral connections. The Slow pathway has high channel capacity while the Fast pathway operates at a fine-grained temporal resolution. We showcase the importance of our two-stream proposal on two diverse datasets: VGG-Sound and EPIC-KITCHENS-100, and achieve state- of-the-art results on both.

Since its introduction in 2018, EPIC-KITCHENS has attracted attention as the largest egocentric video benchmark, offering a unique viewpoint on people's interaction with objects, their attention, and even intention. In this paper, we detail how this large-scale dataset was captured by 32 participants in their native kitchen environments, and densely annotated with actions and object interactions. Our videos depict nonscripted daily activities, as recording is started every time a participant entered their kitchen. Recording took place in four countries by participants belonging to ten different nationalities, resulting in highly diverse kitchen habits and cooking styles. Our dataset features 55 hours of video consisting of 11.5M frames, which we densely labelled for a total of 39.6K action segments and 454.2K object bounding boxes. Our annotation is unique in that we had the participants narrate their own videos (after recording), thus reflecting true intention, and we crowd-sourced ground-truths based on these. We describe our object, action and anticipation challenges, and evaluate several baselines over two test splits, seen and unseen kitchens. We introduce new baselines that highlight the multimodal nature of the dataset and the importance of explicit temporal modelling to discriminate fine-grained actions (e.g., 'closing a tap' from 'opening' it up).

Fine-grained action recognition datasets exhibit environmental bias, where multiple video sequences are captured from a limited number of environments. Training a model in one environment and deploying in another results in a drop in performance due to an unavoidable domain shift. Unsupervised Domain Adaptation (UDA) approaches have frequently utilised adversarial training between the source and target domains. However, these approaches have not explored the multi-modal nature of video within each domain. In this work we exploit the correspondence of modalities as a self-supervised alignment approach for UDA in addition to adversarial alignment (Fig. 1). We test our approach on three kitchens from the large-scale EPIC-Kitchens dataset, using two modalities commonly employed for action recognition: RGB and Optical Flow. We show that multi-modal self-supervision alone improves the performance over source-only training by 2.4% on average. We then combine adversarial training with multi-modal self-supervision, showing that our approach outperforms other UDA methods by 3%.

This work proposes learning to recognize objects from a small number of training examples collected and deployed in-situ. That is, from data collected where the objects are commonly placed or being used, perhaps after first encountering them, the learning algorithm immediately is able to recognize them again. We refer to this method-ology as in-situ learning, and it opposes to the conventional methodology of using complex data acquisition mechanisms, such as rotating tables or synthetic data, to build a large-scale dataset for training convolutional neural networks (ConvNets). To learn in-situ, we propose a novel loss function that generates discriminative features for known and unseen objects, by utilizing a regularization term that reduces the distance between features and their manifold centroid. Additionally, we propose a temporal filter that is particularly useful to quickly react to appearing objects on the scene, which depending on the distance between neighboring video-frame features, it applies a weighted average between the current and the previous frame. Our framework achieves state-of-the-art accuracy for in-situ and on-the-fly learning, for the case of known objects achieves an average increase in accuracy of 3.01%, an increase of 3.3% for novel objects, and an average increase of 7.07% for the combined case, compared with the closest baseline. Utilizing the temporal filtering, led to a further increase in accuracy against nuisances of 7.32% for the known and novels objects case.

Elderly people can be provided with safer and more independent living by the early detection of abnormalities in their performing actions and the frequent assessment of the quality of their motion. Low-cost depth sensing is one of the emerging technologies that can be used for unobtrusive and inexpensive motion abnormality detection and quality assessment. In this study, we develop and evaluate vision-based methods to detect and assess neuromusculoskeletal disorders manifested in common daily activities using three-dimensional skeletal data provided by the SDK of a depth camera (e.g., MS Kinect and Asus Xtion PRO). The proposed methods are based on extracting medically -justified features to compose a simple descriptor. Thereafter, a probabilistic normalcy model is trained on normal motion patterns. For abnormality detection, a test sequence is classified as either normal or abnormal based on its likelihood, which is calculated from the trained normalcy model. For motion quality assessment, a linear regression model is built using the proposed descriptor in order to quantitatively assess the motion quality. The proposed methods were evaluated on four common daily actions-sit to stand, stand to sit, flat walk, and gait on stairs-from two datasets, a publicly released dataset and our dataset that was collected in a clinic from 32 patients suffering from different neuromusculoskeletal disorders and 11 healthy individuals. Experimental results demonstrate promising results, which is a step toward having convenient in-home automatic health care services.

We present a new model to determine relative skill from long videos, through learnable temporal attention modules. Skill determination is formulated as a ranking problem, making it suitable for common and generic tasks. However, for long videos, parts of the video are irrelevant for assessing skill, and there may be variability in the skill exhibited throughout a video. We therefore propose a method which assesses the relative overall level of skill in a long video by attending to its skill-relevant parts. Our approach trains temporal attention modules, learned with only video-level supervision, using a novel rank-aware loss function. In addition to attending to task-relevant video parts, our proposed loss jointly trains two attention modules to separately attend to video parts which are indicative of higher (pros) and lower (cons) skill. We evaluate our approach on the EPIC-Skills dataset and additionally annotate a larger dataset from YouTube videos for skill determination with five previously unexplored tasks. Our method outperforms previous approaches and classic softmax attention on both datasets by over 4% pairwise accuracy, and as much as 12% on individual tasks. We also demonstrate our model’s ability to attend to rank-aware parts of the video.

Recognising actions in videos relies on labelled supervision during training, typically the start and end times of each action instance. This supervision is not only subjective, but also expensive to acquire. Weak video-level supervision has been successfully exploited for recognition in untrimmed videos, however it is challenged when the number of different actions in training videos increases. We propose a method that is supervised by single timestamps located around each action instance, in untrimmed videos. We replace expensive action bounds with sampling distributions initialised from these timestamps. We then use the classifier's response to iteratively update the sampling distributions. We demonstrate that these distributions converge to the location and extent of discriminative action segments. We evaluate our method on three datasets for fine-grained recognition, with increasing number of different actions per video, and show that single timestamps offer a reasonable compromise between recognition performance and labelling effort, performing comparably to full temporal supervision. Our update method improves top-1 test accuracy by up to 5.4%. across the evaluated datasets.

Domain alignment in convolutional networks aims to learn the degree of layer-specific feature alignment beneficial to the joint learning of source and target datasets. While increasingly popular in convolutional networks, there have been no previous attempts to achieve domain alignment in recurrent networks. Similar to spatial features, both source and target domains are likely to exhibit temporal dependencies that can be jointly learnt and aligned. In this paper we introduce Dual-Domain LSTM (DDLSTM), an architecture that is able to learn temporal dependencies from two domains concurrently. It performs cross-contaminated batch normalisation on both input-to-hidden and hidden-to-hidden weights, and learns the parameters for cross-contamination, for both single-layer and multi-layer LSTM architectures. We evaluate DDLSTM on frame-level action recognition using three datasets, taking a pair at a time, and report an average increase in accuracy of 3.5\%. The proposed DDLSTM architecture outperforms standard, fine-tuned, and batch-normalised LSTMs.

We focus on multi-modal fusion for egocentric action recognition, and propose a novel architecture for multimodal temporal-binding, i.e. the combination of modalities within a range of temporal offsets. We train the architecture with three modalities - RGB, Flow and Audio - and combine them with mid-level fusion alongside sparse temporal sampling off used representations. In contrast with previous works, modalities are fused before temporal aggregation, with shared modality and fusion weights over time. Our proposed architecture is trained end-to-end, outperforming individual modalities as well as late-fusion of modalities. We demonstrate the importance of audio in egocentric vision, on per-class basis, for identifying actions as well as interacting objects. Our method achieves state of the art results on both the seen and unseen test sets of the largest egocentric dataset: EPIC-Kitchens, on all metrics using the public leaderboard.

We address the problem of cross-modal fine-grained action retrieval between text and video. Cross-modal retrieval is commonly achieved through learning a shared embedding space, that can indifferently embed modalities. In this paper, we propose to enrich the embedding by disentangling parts-of-speech (PoS) in the accompanying captions. We build a separate multi-modal embedding space for each PoS tag. The outputs of multiple PoS embeddings are then used as input to an integrated multi-modal space, where we perform action retrieval. All embeddings are trained jointly through a combination of PoS-aware and PoS-agnostic losses. Our proposal enables learning specialised embedding spaces that offer multiple views of the same embedded entities. We report the first retrieval results on fine-grained actions for the large-scale EPIC dataset, in a generalised zero-shot setting. Results show the advantage of our approach for both video-to-text and text-to-video action retrieval. We also demonstrate the benefit of disentangling the PoS for the generic task of cross-modal video retrieval on the MSR-VTTdataset.

We address the problem of learning to recognize new objects on-the-fly efficiently. When using CNNs, a typical approach for learning new objects is by fine-tuning the model. However, this approach relies on the assumption that the original training set is available and requires high-end computational resources for training the ever-growing dataset efficiently, which can be unfeasible for robots with limited hardware. To overcome these limitations, we propose a new architecture that: 1) Instead of predicting labels, it learns to generate discriminative and separable embeddings of an object's viewpoints by using a Supervised Triplet Loss, which is easier to implement than current smart mining techniques and the trained model can be applied to unseen objects. 2) Infers an object's identity efficiently by utilizing a lightweight classifier in the features embedding space, this keeps the inference time in the order of milliseconds and can be retrained efficiently when new objects are learned. We evaluate our approach on four real-world images datasets used for Robotics and Computer Vision applications: Amazon Robotics Challenge 2017 by MIT-Princeton, T-LESS, ToyBoX, and CORe50 datasets. Code available at [1].

This paper presents a method for assessing skill from video, applicable to a variety of tasks, ranging from surgery to drawing and rolling pizza dough. We formulate the problem as pairwise (who's better?) and overall (who's best?) ranking of video collections, using supervised deep ranking. We propose a novel loss function that learns discriminative features when a pair of videos exhibit variance in skill, and learns shared features when a pair of videos exhibit comparable skill levels. Results demonstrate our method is applicable across tasks, with the percentage of correctly ordered pairs of videos ranging from 70% to 83% for four datasets. We demonstrate the robustness of our approach via sensitivity analysis of its parameters. We see this work as effort toward the automated organization of how-to video collections and overall, generic skill determination in video.

We present a vision-based trunk-motion tolerant approach which estimates lung volume-time data remotely in forced vital capacity (FVC) and slow vital capacity (SVC) spirometry tests. After temporal modelling of trunk shape, generated using two opposing Kinects in a sequence, the chest-surface respiratory pattern is computed by performing principal component analysis on temporal geometrical features extracted from the chest and posterior shapes. We evaluate our method on a publicly available dataset of 35 subjects (300 sequences) and compare against the state-of-the-art. By filtering complex trunk motions, our proposed method calibrates the entire volume-time data using only the tidal volume scaling factor which reduces the state-of-the-art average normalised L2 error from 0.136 to 0.05.

Automatic discovery of changes in a human's routine is one of the requirements for the future of smart home living, and its contribution to the E-health of the community. In this paper, a Bayesian modelling approach is used which models routine change discovery as a pairwise model selection problem. The method is evaluated on a collected office kitchen dataset that captures snapshots of the routine of the same person over multiple years (2014-2017). The results show that our method is able to detect not only the presence of routine changes, but also which activity patterns have been changed, fully automatically, and in a fully unsupervised manner. Moreover, changes within the same activity pattern can be discovered. Interestingly, discovered changes demonstrate subtle variations that are missed by the visual inspection of a human observer.

Deriving a person's energy expenditure accurately forms the foundation for tracking physical activity levels across many health and lifestyle monitoring tasks. In this study, the authors present a method for estimating calorific expenditure from combined visual and accelerometer sensors by way of an RGB‐Depth camera and a wearable inertial sensor. The proposed individual‐independent framework fuses information from both modalities which leads to improved estimates beyond the accuracy of single modality and manual metabolic equivalents of task (MET) lookup table based methods. For evaluation, the authors introduce a new dataset called SPHERE_RGBD + Inertial_calorie , for which visual and inertial data are simultaneously obtained with indirect calorimetry ground truth measurements based on gas exchange. Experiments show that the fusion of visual and inertial data reduces the estimation error by 8 and 18% compared with the use of visual only and inertial sensor only, respectively, and by 33% compared with a MET‐based approach. The authors conclude from their results that the proposed approach is suitable for home monitoring in a controlled environment.

Manual annotations of temporal bounds for object interactions (i.e. start and end times) are typical training input to recognition, localization and detection algorithms. For three publicly available egocentric datasets, we uncover inconsistencies in ground truth temporal bounds within and across annotators and datasets. We systematically assess the robustness of state-of-the-art approaches to changes in labeled temporal bounds, for object interaction recognition. As boundaries are trespassed, a drop of up to 10% is observed for both Improved Dense Trajectories and Two- Stream Convolutional Neural Network. We demonstrate that such disagreement stems from a limited understanding of the distinct phases of an action, and propose annotating based on the Rubicon Boundaries, inspired by a similarly named cognitive model, for consistent temporal bounds of object interactions. Evaluated on a public dataset, we report a 4% increase in overall accuracy, and an increase in accuracy for 55% of classes when Rubicon Boundaries are used for temporal annotations.

Multiple human tracking (MHT) is a fundamental task in many computer vision applications. Appearance‐based approaches, primarily formulated on RGB data, are constrained and affected by problems arising from occlusions and/or illumination variations. In recent years, the arrival of cheap RGB‐depth devices has led to many new approaches to MHT, and many of these integrate colour and depth cues to improve each and every stage of the process. In this survey, the authors present the common processing pipeline of these methods and review their methodology based (a) on how they implement this pipeline and (b) on what role depth plays within each stage of it. They identify and introduce existing, publicly available, benchmark datasets and software resources that fuse colour and depth data for MHT. Finally, they present a brief comparative evaluation of the performance of those works that have applied their methods to these datasets.

We present an automatic, open source data acquisition and calibration approach using two opposing RGBD sensors (Kinect V2) and demonstrate its efficacy for dynamic object reconstruction in the context of monitoring for remote lung function assessment. First, the relative pose of the two RGBD sensors is estimated through a calibration stage and rigid transformation parameters are computed. These are then used to align and register point clouds obtained from the sensors at frame level. We validated the proposed system by performing experiments on known-size box objects with the results demonstrating accurate measurements. We also report on dynamic object reconstruction by way of human subjects undergoing respiratory functional assessment.

Quantitative assessment of the quality of motion is increasingly in demand by clinicians in healthcare and rehabilitation monitoring of patients. We study and compare the performances of different pose representations and HMM models of dynamics of movement for online quality assessment of human motion. In a general sense, our assessment framework builds a model of normal human motion from skeleton-based samples of healthy individuals. It encapsulates the dynamics of human body pose using robust manifold representation and a first-order Markovian assumption. We then assess deviations from it via a continuous online measure. We compare different feature representations, reduced dimensionality spaces, and HMM models on motions typically tested in clinical settings, such as gait on stairs and flat surfaces, and transitions between sitting and standing. Our dataset is manually labelled by a qualified physiotherapist. The continuous-state HMM, combined with pose representation based on body-joints’ location, outperforms standard discrete-state HMM approaches and other skeleton-based features in detecting gait abnormalities, as well as assessing deviations from the motion model on a frame-by-frame basis.

We present a real-time RGB-D object tracker which manages occlusions and scale changes in a wide variety of scenarios. Its accuracy matches, and in many cases outperforms, state-of-the-art algorithms for precision and it far exceeds most in speed. We build our algorithm on the existing colour-only KCF tracker which uses the `kernel trick' to extend correlation filters for fast tracking. We fuse colour and depth cues as the tracker's features, and furthermore, exploit the depth data to both adjust a given target's scale, and detect and manage occlusions in such a way as to maintain real-time performance, exceeding on average 40~fps. We benchmark our approach using 2 publicly available datasets and make our easy-to-extend modularised code available to other researchers.

Monitoring actions at home can provide essential information for rehabilitation management. This paper presents a comparative study and a dataset for the fully automated, sample-accurate recognition of common home actions in the living room environment using commercial-grade, inexpensive inertial and visual sensors. We investigate the practical home-use of body-worn mobile phone inertial sensors together with an Asus Xmotion RGB-Depth camera to achieve monitoring of daily living scenarios. To test this setup against realistic data, we introduce the challenging SPHERE-H130 action dataset containing 130 sequences of 13 household actions recorded in a home environment. We report automatic recognition results at maximal temporal resolution, which indicate that a vision-based approach outperforms accelerometer provided by two phone-based inertial sensors by an average of 14.85% accuracy for home actions. Further, we report improved accuracy of a vision-based approach over accelerometry on particularly challenging actions as well as when generalising across subjects.

We present a method for the learning and detection of multiple rigid texture-less 3D objects intended to operate at frame rate speeds for video input. The method is geared for fast and scalable learning and detection by combining tractable extraction of edgelet constellations with library lookup based on rotation- and scale-invariant descriptors. The approach learns object views in real-time, and is generative- enabling more objects to be learnt without the need for re-training. During testing, a random sample of edgelet constellations is tested for the presence of known objects. We perform testing of single and multi-object detection on a 30 objects dataset showing detections of any of them within milliseconds from the object’s visibility. The results show the scalability of the approach and its framerate performance. 1

We describe an integrated system for personal workspace monitoring based around an RGB-D sensor. The approach is egocentric, facilitating full flexibility, and operates in real-time, providing object detection and recognition, and 3D trajectory estimation whilst the user undertakes tasks in the workspace. A prototype on-body system developed in the context of work-flow analysis for industrial manipulation and assembly tasks is described. The system is evaluated on two tasks with multiple users, and results indicate that the method is effective, giving good accuracy performance.

This paper presents a complete system on a camera phone that integrates a texture less 3D object detector together with in-situ modelling and tracking. The result is a suite intended for AR applications on the move where objects can be captured, tracked and with the automated detection providing a bridge to either initialize tracking or resume it after measurement loss. The object detector training is online and in-situ and benefits from tight integration with the modelling and tracking processes.

This paper proposes a method for detecting objects carried by pedestrians, such as backpacks and suitcases, from video sequences. In common with earlier work [14], [16] on the same problem, the method produces a representation of motion and shape (known as a temporal template) that has some immunity to noise in foreground segmentations and phase of the walking cycle. Our key novelty is for carried objects to be revealed by comparing the temporal templates against view-specific exemplars generated offline for unencumbered pedestrians. A likelihood map of protrusions, obtained from this match, is combined in a Markov random field for spatial continuity, from which we obtain a segmentation of carried objects using the MAP solution. We also compare the previously used method of periodicity analysis to distinguish carried objects from other protrusions with using prior probabilities for carried-object locations relative to the silhouette. We have reimplemented the earlier state-of-the-art method [14] and demonstrate a substantial improvement in performance for the new method on the PETS2006 data set. The carried-object detector is also tested on another outdoor data set. Although developed for a specific problem, the method could be applied to the detection of irregularities in appearance for other categories of object that move in a periodic fashion.

Recognizing multiple interleaved activities in a video requires implicitly partitioning the detections for each activity. Furthermore, constraints between activities are impor- tant in finding valid explanations for all detections. We use Attribute Multiset Gram- mars (AMGs) as a formal representation for a domain’s knowledge to encode intra- and inter-activity constraints. We show how AMGs can be used to parse all the observa- tions into ‘feasible’ global explanations. We also present an algorithm for building a Bayesian network (BN) given an AMG and a set of detections. The set of labellings of the BN corresponds to the set of all possible parse trees. Finding the best explanation then amounts to finding the maximum a posteriori labeling of the BN. The technique is successfully applied to two different problems including the challenging problem of associating pedestrians and carried objects entering and departing a building.

The ambiguity inherent in a localized analysis of events from video can be resolved by exploiting constraints between events and examining only feasible global explanations. We show how jointly recognizing and linking events can be formulated as labeling of a Bayesian network. The framework can be extended to multiple linking layers, expressing explanations as compositional hierarchies. The best global explanation is the maximum a posteriori (MAP) solution over a set of feasible explanations. The search space is sampled using reversible jump Markov chain Monte Carlo (RJMCMC). We propose a set of general move types that is extensible to multiple layers of linkage, and use simulated annealing to find the MAP solution given all observations. We provide experimental results for a challenging two-layer linkage problem, demonstrating the ability to recognise and link drop and pick events of bicycles in a rack over five days.

Several interesting monitoring applications concern people entering a prescribed area, where they deposit an object in their possession, or collect an object deposited earlier. One example arises in the use of bicycle racks. We propose a novel method for associating each person who deposits an object with the person who later collects it. Our main contribution is to deal with ambiguity in the visual data through the use of global constraints on what is possible. The method is evaluated on a set of practical experiments in a bicycle rack, and applied to online theft detection by comparing the colour profile of associated individuals. 1