Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- David HoggSuggested from co-authorship
- Andrew D. CalwaySuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works197 from public data
- Ego4D: Around the World in 3,000 Hours of Egocentric Video2,124
Ego4D, a massive-scale egocentric video dataset and benchmark suite, is introduced and a host of new benchmark challenges centered around understanding the first-person visual experience in the past, present, and future are presented.
- 2,117
- 2,117
- 1,193
- EPIC-KITCHENS-100806
The pipeline to extend the largest dataset in egocentric vision, EPIC-KITCHENS, using a novel pipeline that allows denser and more complete annotations of fine-grained actions and 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.
- EPIC-KITCHENS-100806
The pipeline to extend the largest dataset in egocentric vision, EPIC-KITCHENS, using a novel pipeline that allows denser and more complete annotations of fine-grained actions and 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.
- Ego-Exo4D: Understanding Skilled Human Activity from First- and Third-Person Perspectives640
To push the frontier of first-person video understanding of skilled human activity, a suite of benchmark tasks and their annotations are presented, including fine-grained activity understanding, proficiency estimation, cross-view translation, and 3D hand/body pose.
- Scaling Egocentric Vision: The "Equation missing" Dataset539
This paper introduces Open image in new window, a large-scale egocentric video benchmark recorded by 32 participants in their native kitchen environments and had the participants narrate their own videos (after recording), thus reflecting true intention, and crowd-sourced ground-truths based on these.
- EPIC-Fusion: Audio-Visual Temporal Binding for Egocentric Action Recognition409
This work proposes a novel architecture for multi-modal temporal-binding, i.e. the combination of modalities within a range of temporal offsets, and demonstrates the importance of audio in egocentric vision, on per-class basis, for identifying actions as well as interacting objects.
- The EPIC-KITCHENS Dataset: Collection, Challenges and Baselines402
This paper details how this large-scale dataset was captured by 32 participants in their native kitchen environments, and densely annotated with actions and object interactions, and introduces new baselines that highlight the multimodal nature of the dataset and the importance of explicit temporal modelling to discriminate fine-grained actions.
- 396
- 279
- Detecting Carried Objects in Short Video Sequences245
A new method for detecting objects such as bags carried by pedestrians depicted in short video sequences by comparing the temporal templates against view-specific exemplars generated offline for unencumbered pedestrians, which yields a segmentation of carried objects using the MAP solution.
- Multi-Modal Domain Adaptation for Fine-Grained Action Recognition241
This work exploits the correspondence of modalities as a self-supervised alignment approach for UDA in addition to adversarial alignment to exploit the multi-modal nature of video within each domain.
- Temporal-Relational CrossTransformers for Few-Shot Action Recognition200
This work proposes a novel approach to few-shot action recognition, finding temporally-corresponding frame tuples between the query and videos in the support set, and outperforms prior work on SSv2 by a wide margin due to the its ability to model temporal relations.
- Fine-Grained Action Retrieval Through Multiple Parts-of-Speech Embeddings191
This paper proposes to enrich the embedding by disentangling parts-of-speech (PoS) in the accompanying captions by building a separate multi-modal embedding space for each PoS tag, which enables learning specialised embedding spaces that offer multiple views of the same embedded entities.
- The Pros and Cons: Rank-Aware Temporal Attention for Skill Determination in Long Videos178
A new model to determine relative skill from long videos, through learnable temporal attention modules, using a novel rank-aware loss function that outperforms previous approaches and classic softmax attention on both datasets by over 4% pairwise accuracy, and as much as 12% on individual tasks.
- 177
- You-Do, I-Learn: Discovering Task Relevant Objects and their Modes of Interaction from Multi-User Egocentric Video166
This work presents a fully unsupervised approach for the discovery of i) task relevant objects and ii) how these objects have been used and introduces the term Mode of Interaction (MOI) to refer to the different ways in which TROs are used.
- 123
- An Outlook into the Future of Egocentric Vision121
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 the authors' every day lives is explored.
- 91
- SPHERE: A Sensor Platform for Healthcare in a Residential Environment88
The home possesses unique characteristics that must be considered in order to develop effective smart home systems that are adopted in the real world and broadly underpinned by shared goals of sustainable development, inclusive user engagement and improved service delivery.
- On Semantic Similarity in Video Retrieval85
This paper proposes several proxies to estimate semantic similarities in large-scale retrieval datasets, without additional annotations, and proposes a move to semantic similarity video retrieval, where multiple videos/captions can be deemed equally relevant, and their relative ranking does not affect a method’s reported performance.
- Real-time RGB-D Tracking with Depth Scaling Kernelised Correlation Filters and Occlusion Handling85
A real-time RGB-D object tracker which manages occlusions and scale changes in a wide variety of scenarios and matches, and in many cases outperforms, state-of-the-art algorithms for precision and it far exceeds most in speed.
- Slow-Fast Auditory Streams for Audio Recognition81
This work learns Slow-Fast auditory streams with separable convolutions and multi-level lateral connections in a two-stream convolutional network for audio recognition, that operates on time-frequency spectrogram inputs.
- A comparative study of pose representation and dynamics modelling for online motion quality78
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.
- A comparative study of pose representation and dynamics modelling for online motion quality assessment78
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.
- 77
- Online quality assessment of human motion from skeleton data71
This work addresses the challenge of analysing the quality of human movements from visual information which has use in a broad range of applications, from diagnosis and rehabilitation to movement optimisation in sports science.
- Real-time Learning and Detection of 3D Texture-less Objects: A Scalable Approach71
This work presents a method for the learning and detection of multiple rigid texture-less 3D objects intended to operate at frame rate speeds for video input by combining tractable extraction of edgelet constellations with library lookup based on rotation-and scale-invariant descriptors.
- Epic-Sounds: A Large-Scale Dataset of Actions that Sound70
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, and proposes an annotation pipeline where annotators temporally label distinguishable audio segments and describe the action that could have caused this sound.
- DS-KCF: a real-time tracker for RGB-D data70
An RGB-D single-object tracker, built upon the extremely fast RGB-only KCF tracker that is able to exploit depth information to handle scale changes, occlusions, and shape changes is proposed.
- Action Recognition From Single Timestamp Supervision in Untrimmed Videos69
This work proposes a method that is supervised by single timestamps located around each action instance, in untrimmed videos, that replaces expensive action bounds with sampling distributions initialised from these timestampeds, and demonstrates that these distributions converge to the location and extent of discriminative action segments.
- 64
- 63
- GenHowTo: Learning to Generate Actions and State Transformations from Instructional Videos59
This work addresses the task of generating temporally consistent and physically plausible images of actions and object state transformations and develops and train a conditioned diffusion model dubbed GenHowTo, which achieves superior performance compared to existing methods.
- With a Little Help from my Temporal Context: Multimodal Egocentric Action Recognition59
A transformer-based multimodal model is proposed that ingests video and audio as input modalities, with an explicit language model providing action sequence context to enhance the predictions.
- You-Do, I-Learn: Egocentric unsupervised discovery of objects and their modes of interaction towards video-based guidance57
An unsupervised approach towards automatically extracting video-based guidance on object usage, from egocentric video and wearable gaze tracking, collected from multiple users while performing tasks is presented.
- Multiple human tracking in RGB‐depth data: a survey54
The authors identify and introduce existing, publicly available, benchmark datasets and software resources that fuse colour and depth data for MHT and present a brief comparative evaluation of the performance of those works that have applied their methods to these datasets.
- Cognitive Learning, Monitoring and Assistance of Industrial Workflows Using Egocentric Sensor Networks54
A scalable concept and an integrated system demonstrator designed for the transfer of learnt workflows from observing multiple expert operators and then transfer the learnt workflow models to novice users are presented.
- 53
- Look-ahead fixations during visuomotor behavior: Evidence from assembling a camping tent47
The diversity of LAFs suggest they may also reflect scene exploration and task relevance, as well as longer term problem solving and task planning, and their frequency was biased to certain objects and subtasks.
- 46
- Rank2Reward: Learning Shaped Reward Functions from Passive Video43
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 is demonstrated and it is demonstrated how the technique can be easily extended to be applicable to web-scale video datasets.
- Action Modifiers: Learning From Adverbs in Instructional Videos43
A method to learn a representation for adverbs from instructional videos using weak supervision from the accompanying narrations, which outperforms all baselines for video-to-adverb retrieval with a performance of 0.719 mAP.
- Toward Respiratory Assessment Using Depth Measurements from a Time-of-Flight Sensor41
Data from a 3-Dimensional time-of-flight camera (found in gaming consoles) used to estimate chest volume during routine spirometry maneuvres are presented to illustrate the potential of 3D time- of-flight cameras to remotely monitor respiratory rate.
- TIM: A Time Interval Machine for Audio-Visual Action Recognition37
The Time Interval Machine (TIM) is proposed where a modality-specific time interval poses as a query to a transformer encoder that ingests a long video input, and can be adapted for action detection, using dense multi-scale inter-val queries, and showing strong performance on the Perception Test.
- Spatial Cognition from Egocentric Video: Out of Sight, Not Out of Mind36
The task of Out of Sight, Not Out of Mind - 3D tracking active objects using observations captured through an egocentric camera is formulated, and it is demonstrated that spatial cognition is critical for correctly locating objects over short and long time scales.
- Trespassing the Boundaries: Labeling Temporal Bounds for Object Interactions in Egocentric Video35
It is demonstrated that disagreement stems from a limited understanding of the distinct phases of an action, and proposed annotating based on the Rubicon Boundaries, inspired by a similarly named cognitive model, for consistent temporal bounds of object interactions is proposed.
- Scaling 4D Representations32
It is shown that by learning from very large video datasets, masked auto-encoding with transformer video models actually scales, consistently improving performance on these 4D tasks, as model size increases from 20M all the way to the largest by far reported self-supervised video model $\unicode{x2013}$ 22B parameters.
- DDLSTM: Dual-Domain LSTM for Cross-Dataset Action Recognition32
The proposed DDLSTM architecture outperforms standard, fine-tuned, and batch-normalised LSTMs, and learns the parameters for cross-contamination, for both single-layer and multi-layer LSTM architectures.
- A comparative home activity monitoring study using visual and inertial sensors31
This paper introduces the challenging SPHERE-H130 action dataset, and reports 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.
- Use Your Head: Improving Long-Tail Video Recognition29
A method is proposed, 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, and employs label mixing to learn robust decision boundaries.
- Remote, Depth-Based Lung Function Assessment29
Experimental results computed over an unprecedented number of clinical patients confirm that chest surface motion is linearly related to the changes in volume of lungs, which establishes the potential toward an accurate, low-cost, and remote alternative to traditional cumbersome methods, such as spirometry.
- What can a cook in Italy teach a mechanic in India? Action Recognition Generalisation Over Scenarios and Locations27
CIR, a method to represent each video as a Cross-Instance Reconstruction of videos from other domains, is proposed and extensive analysis and ablations on ARGO1M show CIR outperforms prior domain generalisation works on all test splits.
- Depth-Based Whole Body Photoplethysmography in Remote Pulmonary Function Testing25
A depth-based whole body photoplethysmography approach, which reduces motion artifacts in depth–time data and highly improves the accuracy of depth- based computed measures, is introduced, which establishes the potential for unconstrained remote respiratory monitoring and diagnosis.
- Estimating visual attention from a head mounted IMU24
This paper evaluates various methods for the estimation of both temporal and spatial visual attention using a head-worn inertial measurement unit (IMU) on a new egocentric dataset from 8 volunteers and compares results with those achievable with a commercial gaze tracker used as ground-truth.
- 21
- The Invisible EgoHand: 3D Hand Forecasting through EgoBody Pose Estimation20
A diffusion-based transformer architecture for Egocentric Hand Forecasting, EgoH4, which takes as input the observation sequence and camera poses, then predicts future 3D motion and poses for both hands of the camera wearer, and leverages full-body pose information, allowing other joints to provide constraints on hand motion.
- The Wisdom of Crowds: Temporal Progressive Attention for Early Action Prediction20
A bottleneck-based attention model that captures the evolution of the action, through progressive sampling over fine-to-coarse scales, called Temporal Progressive (TemPr), which is composed of multiple attention towers, one for each scale.
- 20
- Energy expenditure estimation using visual and inertial sensors19
This study presents a method for estimating calorific expenditure from combined visual and accelerometer sensors by way of an RGB-Depth camera and a wearable inertial sensor and concludes that the proposed approach is suitable for home monitoring in a controlled environment.
- Automated capture and delivery of assistive task guidance with an eyewear computer19
An assistive mixed reality system that aims to augment users in tasks by combining automated and unsupervised information collection with minimally invasive video guides, called GlaciAR (Glass-enabled Contextual Interactions for Augmented Reality).
- Multi-User Egocentric Online System for Unsupervised Assistance on Object Usage19
The paper proposes a method for selecting a suitable video guide to be displayed to a novice user indicating how to use an object, purely triggered by the user’s gaze.
- Learning from One Continuous Video Stream18
Large single-stream learning gains from pre-training with a novel family of future prediction tasks are obtained, and it is found that momentum hurts, and that the pace of weight updates matters, which leads to matching the performance of IID learning with batch size 1, when using the same architecture and without costly replay buffers.
- UnweaveNet: Unweaving Activity Stories18
A video representation explicitly capturing activity threads called a thread bank is introduced, along with a neural controller capable of detecting goal changes and resuming of past activities, together forming UnweaveNet, which is trained and evaluated on sequences from the unscripted egocentric dataset EPIC-KITCHENS.
- 18
- Calorie Counter: RGB-Depth Visual Estimation of Energy Expenditure at Home18
It is concluded from the experiments that the proposed vision pipeline is suitable for home monitoring in a controlled environment, with calorific expenditure estimates above accuracy levels of commonly used manual estimations via METs.
- SEMBED: Semantic Embedding of Egocentric Action Videos18
This work presents SEMBED, an approach for embedding an egocentric object interaction video in a semantic-visual graph to estimate the probability distribution over its potential semantic labels, and shows how it can interpret a challenging dataset of 1225 freely annotated Egocentric videos.
- Who's Better, Who's Best: Skill Determination in Video using Deep Ranking.17
This paper proposes a supervised deep ranking model to learn discriminative features between pairs of videos exhibiting different amounts of skill, and utilises a two-stream Temporal Segment Network to capture both the type and quality of motions and the evolving task state.
- Designing a Video Monitoring System for AAL applications: the SPHERE Case Study17
A detailed description of the hardware and software infrastructure designed and tested in real life scenarios, with particular emphasis on the design considerations employed to foster collaboration, the real time and budget constraints, and mid-scale deployment plan of the case study are presented.
- 16
- 16
- Ego4D: Around the World in 3,600 Hours of Egocentric Video15
Ego4D, a massive-scale egocentric video dataset and benchmark suite, introduced, dramatically expands the volume of diverse egocentric video footage publicly available to the research community and presents a host of new benchmark challenges centered around understanding the first-person visual experience.
- 15
- ShowHowTo: Generating Scene-Conditioned Step-by-Step Visual Instructions14
This work introduces an automatic approach for collecting large step-by-step visual instruction training data from instructional videos and develops and trains ShowHowTo, a video diffusion model capable of generating step-by-step visual instructions consistent with the provided input image.
- EgoPoints: Advancing Point Tracking for Egocentric Videos14
This work introduces EgoPoints, a benchmark for point tracking in egocentric videos, and proposes a pipeline to create semi-real sequences, with automatic ground truth, by combining dynamic Kubric objects with scene points from EPIC Fields.
- Person Re-ID by Fusion of Video Silhouettes and Wearable Signals for Home Monitoring Applications14
A multimodal deep-learning detection framework that maps short video clips and accelerations into a latent space where the Euclidean distance can be measured to match video and acceleration streams and enables home monitoring systems to employ sensor fusion techniques for data analysis.
- Retro-Actions: Learning 'Close' by Time-Reversing 'Open' Videos14
The results show that gestures can be learnt from 'zooming out' in a zero-shot setting, as well as more complex actions with state transitions such as 'digging something out of something' from 'burying something in something'.
- Learning from Streaming Video with Orthogonal Gradients13
This work proposes a geometric modification to standard optimizers, to decorrelate batches by utilising orthogonal gradients during training, and shows the proposed orthogonal optimizer outperforms the strong AdamW in all three scenarios.
- 13
- HOI-Ref: Hand-Object Interaction Referral in Egocentric Vision12
The results demonstrate that VLMs trained for referral on third person images fail to recognise and refer hands and objects in egocentric images, and when fine-tuned on the egocentric HOI-QA dataset, performance improves.
- 12
- 12
- Moment of Untruth: Dealing with Negative Queries in Video Moment Retrieval11
The ability of current SOTA video moment retrieval approaches to adapt to Negative-Aware Video Moment Retrieval is analysed and UniVTG-NA, an adaptation of UniVTG designed to tackle NA-VMR is proposed.
- Get a Grip: Reconstructing Hand-Object Stable Grasps in Egocentric Videos11
The stable grasp definition is developed based on the intuition that the in-contact area between the hand and object should remain stable, and a method to jointly optimise all frames within a stable grasp is proposed, minimising object motions to a latent 1-DoF.
- 11
- 11
- Beyond Action Recognition: Action Completion in RGB-D Data11
This work shows that while features can perform comparably for action recognition, they vary in their ability to recognise incompletion, and test for completion using binary classification on labelled data achieves 95.7% accuracy.
- Personalized Energy Expenditure Estimation: Visual Sensing Approach With Deep Learning10
A vision-based deep learning energy expenditure estimation system for a wide range of daily living activities can be calibrated to a specific person with footage and calorimeter data from 32 seconds of sweeping and 32 Seconds of sitting.
- 10
- 10
- Sit-to-Stand Analysis in the Wild Using Silhouettes for Longitudinal Health Monitoring9
The first fully automated Sit-to-Stand or Stand- to-Sit (StS) analysis framework for long-term monitoring of patients in free-living environments using video silhouettes is presented, highlighting important trends in the recuperation of patients who underwent surgery.
- Weakly-Supervised Completion Moment Detection using Temporal Attention9
This work presents an approach for moment detection from weak video-level labels, that is the moment when the action's goal has been successfully accomplished, and demonstrates how the approach can be used when completion moment supervision is available.
- Hotspots detection for machine operation in egocentric vision9
A good performance of hotspots detection can be achieved by utilizing multiple features, especially touch and hand motion, and detected hotspots in both temporal and spatial locations matches well the positions of key components such as buttons, levers, and other important portions essential for operating the machine.
- Beyond Action Recognition9
This work shows that while features can perform comparably for action recognition, they vary in their ability to recognise incompletion, and test for completion using binary classification on labelled data achieves 95.7% accuracy.
- Explaining Activities as Consistent Groups of Events9
The paper presents a complete framework that starts with a general way to formalise the set of global explanations for a given activity using attribute multiset grammars (AMG), and compares greedy, multiple-hypotheses trees, reversible jump MCMC, and integer programming for finding the Maximum a Posteriori (MAP) solution over the space of explanations.
- Associating People Dropping off and Picking up Objects9
A novel method for associating each person who deposits an object with the person who later collects it is proposed, to deal with ambiguity in the visual data through the use of global constraints on what is possible.
- Egocentric Video-Language Pretraining @ Ego4D Challenge 20228
A pretrained video-language model that is able to transfer its egocentric video-text representation or video-only representation to several video downstream tasks and pioneer Egocentric VLP from pretraining dataset, pretraining objective, and development set is developed.
- Hand-Object Interaction Reasoning8
It is shown that modelling two-handed interactions are critical for action recognition in egocentric video, and that by using positionally-encoded trajectories, the network can better recognise observed interactions.
- 8
- 7
- 7
- TVNet: Temporal Voting Network for Action Localization7
A novel Voting Evidence Module to locate temporal boundaries, more accurately, where temporal contextual evidence is accumulated to predict frame-level probabilities of start and end action boundaries is incorporated within a pipeline to calculate confidence scores and action classes.
- Play Fair: Frame Attributions in Video Models7
This paper proposes a tractable approximation of ESV that scales linearly with the number of frames in the sequence, and offers detailed analysis of supporting/distracting frames, and the relationships of ESVs to the frame's position, class prediction, and sequence length.
- 7
- 7
- Attribute Multiset Grammars for Global Explanations of Activities7
This work uses Attribute Multiset Grammars (AMGs) as a formal representation for a domain’s knowledge to encode intra- and inter-activity constraints to parse all the observations into ‘feasible’ global explanations.
- Perception Test 2024: Challenge Summary and a Novel Hour-Long VideoQA Benchmark6
The Second Perception Test challenge had seven tracks and covered low-level and high-level tasks, with language and non-language interfaces, across video, audio, and text modalities, and introduced a novel video QA benchmark 1h-walk VQA.
- 6
- 6
- Play It Back: Iterative Attention For Audio Recognition5
This work proposes an end-to-end attention-based architecture that through selective repetition attends over the most discriminative sounds across the audio sequence that can consistently achieve state-of-the-art performance across three audio-classification benchmarks.
- Dual-Domain Image Synthesis using Segmentation-Guided GAN5
A segmentation-guided approach to synthesise images that integrate features from two distinct domains, which builds on the successes of few-shot StyleGAN and single-shot semantic segmentation to minimise the amount of training in utilising two domains.
- 5
- 5
- Detecting and Localising Multiple 3D Objects: A Fast and Scalable Approach5
This work presents a generic, scalable and fast framework for concurrently searching multiple rigid textureminimal objects using 2D image edgelet constellations and extends the method to exploit depth information for better clutter removal.
- Context-Aware Multimodal Pretraining4
It is shown that vision-language models can be trained to exhibit significantly increased few-shot adaptation, and equipped with simple, training-free, metric-based adaptation mechanisms, these representations easily surpass more complex and expensive optimization-based schemes, vastly simplifying generalization to new domains.
- Reconstructing Objects along Hand Interaction Timelines in Egocentric Video4
This work proposes to propagate the object's pose along the HIT enabling superior reconstruction using the proposed Constrained Optimisation and Propagation (COP) framework, and focuses on timelines with stable grasps - i.e. where the hand is stably holding an object, effectively maintaining constant contact during use.
- Seeing without Pixels: Perception from Camera Trajectories4
It is found that, contrary to its apparent simplicity, the camera trajectory is a remarkably informative signal to uncover video content, and is established as a lightweight, robust, and versatile modality for perceiving video content.
- No Need for a Lab: Towards Multi-sensory Fusion for Ambient Assisted Living in Real-world Living Homes4
This work improves its architecture and combines it with a tracking functionality that makes it possible to be deployed in real-world homes and shows a novel first example of subject-tailored health monitoring measurement by applying its methodology to a sit-to-stand detector to generate clinically relevant rehabilitation trends.
- 4
- 4
- 4
- It’s Just Another Day: Unique Video Captioning by Discriminitive Prompting3
This paper formulate the problem of unique captioning: Given multiple clips with the same caption, a new caption is generated for each clip that uniquely identifies it, and proposes Captioning by Discriminative Prompting (CDP), which predicts a property that can separate identically captioned clips, and uses it to generate unique captions.
- 3
- Learning Temporal Sentence Grounding From Narrated EgoVideos3
This paper proposes to artificially merge clips to train for temporal grounding in a contrastive manner using text-conditioning attention and this Clip Merging (CliMer) approach is shown to be effective when compared with a high performing TSG method.
- Centre Stage: Centricity-based Audio-Visual Temporal Action Detection3
This paper proposes a novel network head to estimate the closeness of timesteps to the action centre, which is called the centricity score, and demonstrates the correlation between the distance from the timestep to theaction centre and the accuracy of the predicted boundaries.
- Semantically Selective Augmentation for Deep Compact Person Re-Identification3
This work proposes to augment limited training data via sampling from a deep convolutional generative adversarial network (DCGAN), whose discriminator is constrained by a semantic classifier to explicitly control the domain specificity of the generation process.
- 3
- 3
- 3
- Real-time Estimation of Physical Activity Intensity for Daily Living3
The proposed real-time vision pipeline is suitable for monitoring physical activity levels in a controlled environment with higher accuracy than the commonly used manual estimation via metabolic lookup tables (METs), whilst being significantly faster than existing automated methods.
- 3
- 3
- 3
- 3
- Interactive image-based model building for handheld devices3
A system is extended by incorporating an object detection process which starts and/or re-initializes the tracker when lost creating a more fluid interaction process and comparing two interactive methods to specify 3D vertices.
- PointSt3R: Point Tracking through 3D Grounded Correspondence2
This paper demonstrates that DUSt3R and MASt3R are competitive point trackers when focusing on static points, and proposes 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.
- Improving and Evaluating Hand-Object Interaction Detection2
Hoi-DETR is a new framework that introduces hand-object and object-object interactions to the Co-DETR architecture to produce a state-of-the-art method for HOI understanding and ablations confirm the contributions of each model component.
- Unique Lives, Shared World: Learning from Single-Life Videos2
The "single-life"learning paradigm is introduced, where a distinct vision model exclusively on egocentric videos captured by one individual is trained, establishing that the shared structure of the world, both leads to consistency in models trained on individual lives, and provides a powerful signal for visual representation learning.
- 2
- Refining Action Boundaries for One-stage Detection2
This work incorporates the estimation of boundary confidence into one-stage anchor-free detection, through an additional prediction head that predicts the refined boundaries with higher confidence, to obtain state-of-the-art performance on the challenging EPICKITCHENS-100 action detection as well as the standard THUMOS14 action detection benchmarks.
- Meta-Learning with Context-Agnostic Initialisations2
This work incorporates a context-adversarial component into the meta-learning process, which produces an initialisation for fine-tuning to target which is both context-agnostic and task-generalised.
- Integration of Experts' and Beginners' Machine Operation Experiences to Obtain a Detailed Task Model2
A novel framework for integrating beginners’ machine operational experiences with those of experts’ to obtain a detailed task model is proposed and good potential for modeling and obtaining important properties of the task is demonstrated.
- 2
- 2
- 2
- P83 Remote pulmonary function testing – computer gaming in the respiratory world2
A pilot data is described from the initial development of a new technique for non-invasively assessing lung volume and pulmonary function measurements using a 3D time-of-flight depth camera similar to those found in many home gaming consoles.
- Perception Test 2025: Challenge Summary and a Unified VQA Extension1
The Third Perception Test challenge was organised as a full-day workshop alongside the IEEE/CVF International Conference on Computer Vision (ICCV) 2025, and introduced a novel subset that reformulates traditional perception tasks as multiple-choice video QA questions that video-language models can natively tackle.
- Prime and Reach: Synthesising Body Motion for Gaze-Primed Object Reach1
This work curates 23.7K gaze-primed human motion sequences for reaching target object locations from five publicly available datasets, and evaluates the ability of the generated motion to imitate natural human movement through several metrics, including the 'Reach Success' and a newly introduced'Prime Success'metric'.
- Active Object Segmentation: A New Modality for Egocentric Action Recognition1
An indirect method for obtaining RGB-like representations from segmentations via a generative model and a direct method that directly learns presence, relative positioning, and motion representations from segmentations via a multi-stream network are presented.
- 1
- Inertial Hallucinations -- When Wearable Inertial Devices Start Seeing Things1
This work proposes a novel approach to multimodal sensor fusion for Ambient Assisted Living (AAL) which takes advantage of learning using privileged information (LUPI), and fuses the concept of modality hallucination with triplet learning to train a model with different modalities to handle missing sensors at inference time.
- Human-Centric Object Interactions - A Fine-Grained Perspective from Egocentric Videos1
A fine(r)-grained perspective onto human-object interactions from video is argued for, using multi-modal footage (appearance, motion, audio, language) and approaches for determining skill or expertise from video sequences are presented.
- 1
- 1
- 1
- 1
- 1
- 1
- –
- Beyond Caption-Based Queries for Video Moment Retrieval–
This work investigates the degradation of existing VMR methods, particularly of DETR architectures, when trained on caption-based queries but evaluated on search queries, and identifies a critical issue in these architectures -- an active decoder-query collapse -- as a primary cause of the poor generalization to multi-moment instances.
- –
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?–
This work proposes a new learning paradigm that combines hand-object masked training, which enables robust reasoning from partial hand or object observations, and an HOI-dynamics-aware decoder that explicitly learns hand- and object-centric embeddings through auxiliary predictions of their locations and semantics, enhancing sensitivity to both cues.
- –
- –
- –
- –
- Segmenting Collision Sound Sources in Egocentric Videos–
This work proposes a weakly-supervised method for audio-conditioned segmentation, utilising foundation models (CLIP and SAM2) and incorporates egocentric cues, i.e. objects in hands, to find acting objects that can potentially be collision sound sources.
- The N-Body Problem: Parallel Execution from Single-Person Egocentric Video–
A structured prompting strategy is introduced that guides a Vision-Language Model (VLM) to reason about the 3D environment, object usage, and temporal dependencies, producing a viable parallel execution.
- –
- –
- –
- –
- An Evaluation of OCR on Egocentric Data–
It is demonstrated that existing OCR methods struggle with rotated text, which is frequently observed on objects being handled, and a simple rotate-and-merge procedure is introduced which can be applied to pre-trained OCR models that halves the normalized edit distance error.
- –
- –
- –
- –
- –
- Deep Compact Person Re-Identification with Distractor Synthesis via Guided DC-GANs–
It is concluded that whilst re-identification performance is highly sensitive to the structure of datasets, distractor augmentation and network compression have a role to play for enhancing performance characteristics for larger scale applications.
- –
- –
- –
- Instance-level Object Recognition Using Deep Temporal Coherence–
This paper evaluates the use of temporal filters such as Cumulative Moving Average and a machine learning approach using Recurrent Neural Networks for this task of instance object recognition and shows that by exploiting temporal coherence, models trained with a few data points perform comparably to when the whole dataset is available.
- Shared Use Routes for People Who Walk or Cycle: Addressing the Challenges–
A social media campaign encouraged pedestrians and other users to pinpoint shared spaces on a map of Bristol and insert a comment about the shared use, and five key themes emerged: space, speed, surfaces, signage and social behaviour.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.