Andrew Gilbert

Mona Ahmadian, Frank Guerin, Andrew Gilbert

Self-supervised learning (SSL) techniques have recently produced outstanding results in learning visual representations from unlabeled videos. Despite the importance of motion in supervised learning techniques for action recognition, SSL methods often do not explicitly consider motion information in videos. To address this issue, we propose MOFO (MOtion FOcused), a novel SSL method for focusing representation learning on the motion area of a video, for action recognition. MOFO automatically detects motion areas in videos and uses these to guide the self-supervision task. We use a masked autoencoder which randomly masks out a high proportion of the input sequence; we force a specified percentage of the inside of the motion area to be masked and the remainder from outside. We further incorporate motion information into the finetuning step to emphasise motion in the downstream task. We demonstrate that our motion-focused innovations can significantly boost the performance of the currently leading SSL method (VideoMAE) for action recognition. Our method improves the recent self-supervised Vision Transformer (ViT), VideoMAE, by achieving +2.6%, +2.1%, +1.3% accuracy on Epic-Kitchens verb, noun and action classification, respectively, and +4.7% accuracy on Something-Something V2 action classification. Our proposed approach significantly improves the performance of the current SSL method for action recognition, indicating the importance of explicitly encoding motion in SSL.

Xu Dong, Xinran Liu, Wanqing Li et al.

Long-term Action Quality Assessment (AQA) evaluates the execution of activities in videos. However, the length presents challenges in fine-grained interpretability, with current AQA methods typically producing a single score by averaging clip features, lacking detailed semantic meanings of individual clips. Long-term videos pose additional difficulty due to the complexity and diversity of actions, exacerbating interpretability challenges. While query-based transformer networks offer promising long-term modeling capabilities, their interpretability in AQA remains unsatisfactory due to a phenomenon we term Temporal Skipping, where the model skips self-attention layers to prevent output degradation. To address this, we propose an attention loss function and a query initialization method to enhance performance and interpretability. Additionally, we introduce a weight-score regression module designed to approximate the scoring patterns observed in human judgments and replace conventional single-score regression, improving the rationality of interpretability. Our approach achieves state-of-the-art results on three real-world, long-term AQA benchmarks. Our code is available at: https://github.com/dx199771/Interpretability-AQA

Edward Fish, Jon Weinbren, Andrew Gilbert

Temporal Action Localization (TAL) aims to identify actions' start, end, and class labels in untrimmed videos. While recent advancements using transformer networks and Feature Pyramid Networks (FPN) have enhanced visual feature recognition in TAL tasks, less progress has been made in the integration of audio features into such frameworks. This paper introduces the Multi-Resolution Audio-Visual Feature Fusion (MRAV-FF), an innovative method to merge audio-visual data across different temporal resolutions. Central to our approach is a hierarchical gated cross-attention mechanism, which discerningly weighs the importance of audio information at diverse temporal scales. Such a technique not only refines the precision of regression boundaries but also bolsters classification confidence. Importantly, MRAV-FF is versatile, making it compatible with existing FPN TAL architectures and offering a significant enhancement in performance when audio data is available.

Anastasia Anichenko, Frank Guerin, Andrew Gilbert

We investigate a human-like interpretable model of video understanding. Humans recognise complex activities in video by recognising critical spatio-temporal relations among explicitly recognised objects and parts, for example, an object entering the aperture of a container. To mimic this we build on a model which uses positions of objects and hands, and their motions, to recognise the activity taking place. To improve this model we focussed on three of the most confused classes (for this model) and identified that the lack of 3D information was the major problem. To address this we extended our basic model by adding 3D awareness in two ways: (1) A state-of-the-art object detection model was fine-tuned to determine the difference between "Container" and "NotContainer" in order to integrate object shape information into the existing object features. (2) A state-of-the-art depth estimation model was used to extract depth values for individual objects and calculate depth relations to expand the existing relations used our interpretable model. These 3D extensions to our basic model were evaluated on a subset of three superficially similar "Putting" actions from the Something-Something-v2 dataset. The results showed that the container detector did not improve performance, but the addition of depth relations made a significant improvement to performance.

Adrienne Deganutti, Simon Hadfield, Andrew Gilbert

Audio Description is a narrated commentary designed to aid vision-impaired audiences in perceiving key visual elements in a video. While short-form video understanding has advanced rapidly, a solution for maintaining coherent long-term visual storytelling remains unresolved. Existing methods rely solely on frame-level embeddings, effectively describing object-based content but lacking contextual information across scenes. We introduce DANTE-AD, an enhanced video description model leveraging a dual-vision Transformer-based architecture to address this gap. DANTE-AD sequentially fuses both frame and scene level embeddings to improve long-term contextual understanding. We propose a novel, state-of-the-art method for sequential cross-attention to achieve contextual grounding for fine-grained audio description generation. Evaluated on a broad range of key scenes from well-known movie clips, DANTE-AD outperforms existing methods across traditional NLP metrics and LLM-based evaluations.

Edward Fish, Andrew Gilbert

Few-shot temporal action localization (TAL) methods that adapt large models via single-prompt tuning often fail to produce precise temporal boundaries. This stems from the model learning a non-discriminative mean representation of an action from sparse data, which compromises generalization. We address this by proposing a new paradigm based on multi-prompt ensembles, where a set of diverse, learnable prompts for each action is encouraged to specialize on compositional sub-events. To enforce this specialization, we introduce PLOT-TAL, a framework that leverages Optimal Transport (OT) to find a globally optimal alignment between the prompt ensemble and the video's temporal features. Our method establishes a new state-of-the-art on the challenging few-shot benchmarks of THUMOS'14 and EPIC-Kitchens, without requiring complex meta-learning. The significant performance gains, particularly at high IoU thresholds, validate our hypothesis and demonstrate the superiority of learning distributed, compositional representations for precise temporal localization.

Mona Ahmadian, Amir Shirian, Frank Guerin et al.

Real-world videos often contain overlapping events and complex temporal dependencies, making multimodal interaction modeling particularly challenging. We introduce DEL, a framework for dense semantic action localization, aiming to accurately detect and classify multiple actions at fine-grained temporal resolutions in long untrimmed videos. DEL consists of two key modules: the alignment of audio and visual features that leverage masked self-attention to enhance intra-mode consistency and a multimodal interaction refinement module that models cross-modal dependencies across multiple scales, enabling high-level semantics and fine-grained details. Our method achieves state-of-the-art performance on multiple real-world Temporal Action Localization (TAL) datasets, UnAV-100, THUMOS14, ActivityNet 1.3, and EPIC-Kitchens-100, surpassing previous approaches with notable average mAP gains of +3.3%, +2.6%, +1.2%, +1.7% (verb), and +1.4% (noun), respectively.

Kar Balan, Andrew Gilbert, John Collomosse

The rise of Generative AI (GenAI) has sparked significant debate over balancing the interests of creative rightsholders and AI developers. As GenAI models are trained on vast datasets that often include copyrighted material, questions around fair compensation and proper attribution have become increasingly urgent. To address these challenges, this paper proposes a framework called Content ARCs (Authenticity, Rights, Compensation). By combining open standards for provenance and dynamic licensing with data attribution, and decentralized technologies, Content ARCs create a mechanism for managing rights and compensating creators for using their work in AI training. We characterize several nascent works in the AI data licensing space within Content ARCs and identify where challenges remain to fully implement the end-to-end framework.

Mona Ahmadian, Frank Guerin, Andrew Gilbert

This paper demonstrates a self-supervised approach for learning semantic video representations. Recent vision studies show that a masking strategy for vision and natural language supervision has contributed to developing transferable visual pretraining. Our goal is to achieve a more semantic video representation by leveraging the text related to the video content during the pretraining in a fully self-supervised manner. To this end, we present FILS, a novel self-supervised video Feature prediction In semantic Language Space (FILS). The vision model can capture valuable structured information by correctly predicting masked feature semantics in language space. It is learned using a patch-wise video-text contrastive strategy, in which the text representations act as prototypes for transforming vision features into a language space, which are then used as targets for semantically meaningful feature prediction using our masked encoder-decoder structure. FILS demonstrates remarkable transferability on downstream action recognition tasks, achieving state-of-the-art on challenging egocentric datasets, like Epic-Kitchens, Something-SomethingV2, Charades-Ego, and EGTEA, using ViT-Base. Our efficient method requires less computation and smaller batches compared to previous works.

Joseph Chrol-Cannon, Andrew Gilbert, Ranko Lazic et al.

Video activity recognition by deep neural networks is impressive for many classes. However, it falls short of human performance, especially for challenging to discriminate activities. Humans differentiate these complex activities by recognising critical spatio-temporal relations among explicitly recognised objects and parts, for example, an object entering the aperture of a container. Deep neural networks can struggle to learn such critical relationships effectively. Therefore we propose a more human-like approach to activity recognition, which interprets a video in sequential temporal phases and extracts specific relationships among objects and hands in those phases. Random forest classifiers are learnt from these extracted relationships. We apply the method to a challenging subset of the something-something dataset and achieve a more robust performance against neural network baselines on challenging activities.

Dan Ruta, Saeid Motiian, Baldo Faieta et al.

We present ALADIN (All Layer AdaIN); a novel architecture for searching images based on the similarity of their artistic style. Representation learning is critical to visual search, where distance in the learned search embedding reflects image similarity. Learning an embedding that discriminates fine-grained variations in style is hard, due to the difficulty of defining and labelling style. ALADIN takes a weakly supervised approach to learning a representation for fine-grained style similarity of digital artworks, leveraging BAM-FG, a novel large-scale dataset of user generated content groupings gathered from the web. ALADIN sets a new state of the art accuracy for style-based visual search over both coarse labelled style data (BAM) and BAM-FG; a new 2.62 million image dataset of 310,000 fine-grained style groupings also contributed by this work.

Edward Fish, Jon Weinbren, Andrew Gilbert

Movie genre classification is an active research area in machine learning. However, due to the limited labels available, there can be large semantic variations between movies within a single genre definition. We expand these 'coarse' genre labels by identifying 'fine-grained' semantic information within the multi-modal content of movies. By leveraging pre-trained 'expert' networks, we learn the influence of different combinations of modes for multi-label genre classification. Using a contrastive loss, we continue to fine-tune this 'coarse' genre classification network to identify high-level intertextual similarities between the movies across all genre labels. This leads to a more 'fine-grained' and detailed clustering, based on semantic similarities while still retaining some genre information. Our approach is demonstrated on a newly introduced multi-modal 37,866,450 frame, 8,800 movie trailer dataset, MMX-Trailer-20, which includes pre-computed audio, location, motion, and image embeddings.

Kary Ho, Andrew Gilbert, Hailin Jin et al.

We present a neural architecture search (NAS) technique to enhance the performance of unsupervised image de-noising, in-painting and super-resolution under the recently proposed Deep Image Prior (DIP). We show that evolutionary search can automatically optimize the encoder-decoder (E-D) structure and meta-parameters of the DIP network, which serves as a content-specific prior to regularize these single image restoration tasks. Our binary representation encodes the design space for an asymmetric E-D network that typically converges to yield a content-specific DIP within 10-20 generations using a population size of 500. The optimized architectures consistently improve upon the visual quality of classical DIP for a diverse range of photographic and artistic content.

Andrew Gilbert, Matthew Trumble, Adrian Hilton et al.

We aim to simultaneously estimate the 3D articulated pose and high fidelity volumetric occupancy of human performance, from multiple viewpoint video (MVV) with as few as two views. We use a multi-channel symmetric 3D convolutional encoder-decoder with a dual loss to enforce the learning of a latent embedding that enables inference of skeletal joint positions and a volumetric reconstruction of the performance. The inference is regularised via a prior learned over a dataset of view-ablated multi-view video footage of a wide range of subjects and actions, and show this to generalise well across unseen subjects and actions. We demonstrate improved reconstruction accuracy and lower pose estimation error relative to prior work on two MVV performance capture datasets: Human 3.6M and TotalCapture.

Andrew Gilbert, Marco Volino, John Collomosse et al.

We present a convolutional autoencoder that enables high fidelity volumetric reconstructions of human performance to be captured from multi-view video comprising only a small set of camera views. Our method yields similar end-to-end reconstruction error to that of a probabilistic visual hull computed using significantly more (double or more) viewpoints. We use a deep prior implicitly learned by the autoencoder trained over a dataset of view-ablated multi-view video footage of a wide range of subjects and actions. This opens up the possibility of high-end volumetric performance capture in on-set and prosumer scenarios where time or cost prohibit a high witness camera count.

Andrew Gilbert, Richard Bowden

Within the field of image and video recognition, the traditional approach is a dataset split into fixed training and test partitions. However, the labelling of the training set is time-consuming, especially as datasets grow in size and complexity. Furthermore, this approach is not applicable to the home user, who wants to intuitively group their media without tirelessly labelling the content. Our interactive approach is able to iteratively cluster classes of images and video. Our approach is based around the concept of an image signature which, unlike a standard bag of words model, can express co-occurrence statistics as well as symbol frequency. We efficiently compute metric distances between signatures despite their inherent high dimensionality and provide discriminative feature selection, to allow common and distinctive elements to be identified from a small set of user labelled examples. These elements are then accentuated in the image signature to increase similarity between examples and pull correct classes together. By repeating this process in an online learning framework, the accuracy of similarity increases dramatically despite labelling only a few training examples. To demonstrate that the approach is agnostic to media type and features used, we evaluate on three image datasets (15 scene, Caltech101 and FG-NET), a mixed text and image dataset (ImageTag), a dataset used in active learning (Iris) and on three action recognition datasets (UCF11, KTH and Hollywood2). On the UCF11 video dataset, the accuracy is 86.7% despite using only 90 labelled examples from a dataset of over 1200 videos, instead of the standard 1122 training videos. The approach is both scalable and efficient, with a single iteration over the full UCF11 dataset of around 1200 videos taking approximately 1 minute on a standard desktop machine.