Abstract Unlike the sparse label action detection task, where a single action occurs in each timestamp of a video, in a dense multi-label scenario, actions can overlap temporally. To address this challenging task, it is necessary to simultaneously learn (i) co-occurrence action relationships and (ii) temporal dependencies. Current methods model co-occurrence action relationships by explicitly embedding class relations into the transformer network architecture. However, these approaches are not computationally efficient, as the network needs to compute all possible pair action class relations. In this paper, we overcome this by introducing a novel framework trained through a novel learning paradigm that allows the network to benefit from explicitly modelling temporal co-occurrence action dependencies during training without imposing their computational overhead during inference. Furthermore, to model temporal information, recent approaches extract multi-scale temporal features through hierarchical transformer-based networks. However, the self-attention mechanism in transformers inherently loses temporal positional information. We argue that combining this with multiple sub-sampling processes in hierarchical designs can lead to further loss of positional information. Preserving this information is essential for accurate action detection. In this paper, we address this issue by proposing a novel transformer network that (a) employs a non-hierarchical structure when modelling different ranges of temporal dependencies and (b) embeds relative positional encoding in its transformer layers. We evaluate the performance of our proposed approach on two challenging dense multi-label benchmark datasets and show that our method improves the current state-of-the-art results by 1.1% and 0.6% per-frame mAP on the Charades and MultiTHUMOS datasets, respectively, achieving new state-of-the-art per-frame mAP results at 26.5% and 44.6%, respectively. We also performed extensive ablation studies to examine the impact of the different components of our proposed approach. Our code will be released upon paper publication

Self-supervised pre-trained audio networks have seen widespread adoption in real-world systems, particularly in multi-modal large language models. These networks are often employed in a frozen state, under the assumption that the self-supervised pre-training has sufficiently equipped them to handle real-world audio. However, a critical question remains: how well do these models actually perform in real-world conditions, where audio is typically polyphonic and complex, involving multiple overlapping sound sources? Current audio self-supervised learning (SSL) methods are often benchmarked on datasets predominantly featuring monophonic audio, such as environmental sounds, and speech. As a result, the ability of SSL models to generalize to polyphonic audio, a common characteristic in natural scenarios, remains underexplored. This limitation raises concerns about the practical robustness of SSL models in more realistic audio settings. To address this gap, we introduce Self-Supervised Learning from Audio Mixtures (SSLAM), a novel direction in audio SSL research, designed to improve the model’s ability to learn from polyphonic data while maintaining strong performance on monophonic data. We thoroughly evaluate SSLAM on standard audio SSL benchmark datasets which are predominantly monophonic and conduct a comprehensive comparative analysis against state-of-the-art (SOTA) methods using a range of high-quality, publicly available polyphonic datasets. SSLAM not only improves model performance on polyphonic audio, but also maintains or exceeds performance on standard audio SSL benchmarks. Notably, it achieves up to a 3.9% improvement on the AudioSet-2M(AS-2M), reaching a mean average precision (mAP) of 50.2. For polyphonic datasets, SSLAM sets new SOTA in both linear evaluation and fine-tuning regimes with performance improvements of up to 9.1%(mAP). These results demonstrate SSLAM's effectiveness in both polyphonic and monophonic soundscapes, significantly enhancing the performance of audio SSL models. Code and pre-trained models are available at https://github.com/ta012/SSLAM.

Existing video captioning benchmarks and models lack causal-temporal narrative, which is sequences of events linked through cause and effect, unfolding over time and driven by characters or agents. This lack of narrative restricts models’ ability to generate text descriptions that capture the causal and temporal dynamics inherent in video content. To address this gap, we propose NarrativeBridge, an approach comprising of: (1) a novel Causal-Temporal Narrative (CTN) captions benchmark generated using a large language model and few-shot prompting, explicitly encoding cause-effect temporal relationships in video descriptions; and (2) a Cause-Effect Network (CEN) with separate encoders for capturing cause and effect dynamics, enabling effective learning and generation of captions with causal-temporal narrative. Extensive experiments demonstrate that CEN significantly outperforms state-of-the-art models in articulating the causal and temporal aspects of video content: 17.88 and 17.44 CIDEr on the MSVD-CTN and MSRVTT-CTN datasets, respectively. Cross-dataset evaluations further showcase CEN’s strong generalization capabilities. The proposed framework understands and generates nuanced text descriptions with intricate causal-temporal narrative structures present in videos, addressing a critical limitation in video captioning. For project details, visit https://narrativebridge.github.io/.

Convolutional neural networks (CNNs) and Transformer-based networks have recently enjoyed significant attention for various audio classification and tagging tasks following their wide adoption in the computer vision domain. Despite the difference in information distribution between audio spectrograms and natural images, there has been limited exploration of effective information retrieval from spectrograms using domain-specific layers tailored for the audio domain. In this paper, we leverage the power of the Multi-Axis Vision Transformer (MaxViT) to create DTF-AT (Decoupled Time-Frequency Audio Transformer) that facilitates interactions across time, frequency, spatial, and channel dimensions. The proposed DTF-AT architecture is rigorously evaluated across diverse audio and speech classification tasks, consistently establishing new benchmarks for state-of-the-art (SOTA) performance. Notably, on the challenging AudioSet 2M classification task, our approach demonstrates a substantial improvement of 4.4% when the model is trained from scratch and 3.2% when the model is initialised from ImageNet-1K pretrained weights. In addition, we present comprehensive ablation studies to investigate the impact and efficacy of our proposed approach. The codebase and pretrained weights are available on https://github.com/ta012/DTFAT.git

In the domain of audio transformer architectures, prior research has extensively investigated isotropic architectures that capture the global context through full self-attention and hierarchical architectures that progressively transition from local to global context utilising hierarchical structures with convolutions or window-based attention. However, the idea of imbuing each individual block with both local and global contexts, thereby creating a hybrid transformer block, remains relatively under-explored in the field.To facilitate this exploration, we introduce Multi Axis Audio Spectrogram Transformer (Max-AST), an adaptation of MaxViT to the audio domain. Our approach leverages convolution, local window-attention, and global grid-attention in all the transformer blocks. The proposed model excels in efficiency compared to prior methods and consistently outperforms state-of-the-art techniques, achieving significant gains of up to 2.6% on the AudioSet full set. Further, we performed detailed ablations to analyse the impact of each of these components on audio feature learning. The source code is available at https://github.com/ta012/MaxAST.git

Leveraging machine learning techniques, in the context of object-based media production, could enable provision of personalized media experiences to diverse audiences. To fine-tune and evaluate techniques for personalization applications, as well as more broadly, datasets which bridge the gap between research and production are needed. We introduce and release such a dataset, themed around a UK weather forecast and shot against a blue-screen background, of three professional actors/presenters – one male and one female (English) and one female (British Sign Language). Scenes include both production and research-oriented examples, with a range of dialogues and actions. Capture techniques consisted of a synchronized 4K resolution 16-camera array, production-typical microphones plus professional audio mix, a 16-channel microphone array with collocated Grasshopper3 camera, and a photogrammetry array. We demonstrate applications relevant to virtual production and creation of personalized media including neural radiance fields, shadow casting, action/event detection, speaker source tracking and video captioning.

In the context of Audio Visual Question Answering (AVQA) tasks, the audio and visual modalities could be learnt on three levels: 1) Spatial, 2) Temporal, and 3) Semantic. Existing AVQA methods suffer from two major shortcomings; the audio-visual (AV) information passing through the network isn’t aligned on Spatial and Temporal levels; and, inter-modal (audio and visual) Semantic information is often not balanced within a context; this results in poor performance. In this paper, we propose a novel end-to-end Contextual Multi-modal Alignment (CAD) network that addresses the challenges in AVQA methods by i) introducing a parameter-free stochastic Contextual block that ensures robust audio and visual alignment on the Spatial level; ii) proposing a pre-training technique for dynamic audio and visual alignment on Temporal level in a self-supervised setting, and iii) introducing a cross-attention mechanism to balance audio and visual information on Semantic level. The proposed novel CAD network improves the overall performance over the state-of-the-art methods on average by 9.4% on the MUSIC-AVQA dataset. We also demonstrate that our proposed contributions to AVQA can be added to the existing methods to improve their performance without additional complexity requirements.

We introduce the first method to automatically generate 3D mesh sequences from text, inspired by the challenging problem of Sign Language Production (SLP). The approach only requires simple 2D annotations for training, which can be automatically extracted from video. Rather than incorporating high-definition or motion capture data, we propose back-translation as a powerful paradigm for supervision: By first addressing the arguably simpler problem of translating 2D pose sequences to text, we can leverage this to drive a transformer-based architecture to translate text to 2D poses. These are then used to drive a 3D mesh generator. Our mesh generator Pose2Mesh uses temporal information, to enforce temporal coherence and significantly reduce processing time. The approach is evaluated by generating 2D pose, and 3D mesh sequences in DGS (German Sign Language) from German language sentences. An extensive analysis of the approach and its sub-networks is conducted, reporting BLEU and ROUGE scores, as well as Mean 2D Joint Distance. Our proposed Text2Pose model outperforms the current state-of-the-art in SLP, and we establish the first benchmark for the complex task of text-to-3D-mesh-sequence generation with our Text2Mesh model.

Abstract We introduce the first approach to solve the challenging problem of automatic 4D visual scene understanding for complex dynamic scenes with multiple interacting people from multi-view video. Our approach simultaneously estimates a detailed model that includes a per-pixel semantically and temporally coherent reconstruction, together with instance-level segmentation exploiting photo-consistency, semantic and motion information. We further leverage recent advances in 3D pose estimation to constrain the joint semantic instance segmentation and 4D temporally coherent reconstruction. This enables per person semantic instance segmentation of multiple interacting people in complex dynamic scenes. Extensive evaluation of the joint visual scene understanding framework against state-of-the-art methods on challenging indoor and outdoor sequences demonstrates a significant ( $$\approx 40\%$$ ≈ 40 % ) improvement in semantic segmentation, reconstruction and scene flow accuracy. In addition to the evaluation on several indoor and outdoor scenes, the proposed joint 4D scene understanding framework is applied to challenging outdoor sports scenes in the wild captured with manually operated wide-baseline broadcast cameras.

We present a new end-to-end learning framework to obtain detailed and spatially coherent reconstructions of multiple people from a single image. Existing multi-person methods suffer from two main drawbacks: they are often model-based and therefore cannot capture accurate 3D models of people with loose clothing and hair; or they require manual intervention to resolve occlusions or interactions. Our method addresses both limitations by introducing the first end-to-end learning approach to perform model-free implicit reconstruction for realistic 3D capture of multiple clothed people in arbitrary poses (with occlusions) from a single image. Our network simultaneously estimates the 3D geometry of each person and their 6DOF spatial locations, to obtain a coherent multi-human reconstruction. In addition, we introduce a new synthetic dataset that depicts images with a varying number of inter-occluded humans and a variety of clothing and hair styles. We demonstrate robust, high-resolution reconstructions on images of multiple humans with complex occlusions, loose clothing and a large variety of poses and scenes. Our quantitative evaluation on both synthetic and real world datasets demonstrates state-of-the-art performance with significant improvements in the accuracy and completeness of the reconstructions over competing approaches.

Abstract Existing techniques for dynamic scene reconstruction from multiple wide-baseline cameras primarily focus on reconstruction in controlled environments, with fixed calibrated cameras and strong prior constraints. This paper introduces a general approach to obtain a 4D representation of complex dynamic scenes from multi-view wide-baseline static or moving cameras without prior knowledge of the scene structure, appearance, or illumination. Contributions of the work are: an automatic method for initial coarse reconstruction to initialize joint estimation; sparse-to-dense temporal correspondence integrated with joint multi-view segmentation and reconstruction to introduce temporal coherence; and a general robust approach for joint segmentation refinement and dense reconstruction of dynamic scenes by introducing shape constraint. Comparison with state-of-the-art approaches on a variety of complex indoor and outdoor scenes, demonstrates improved accuracy in both multi-view segmentation and dense reconstruction. This paper demonstrates unsupervised reconstruction of complete temporally coherent 4D scene models with improved non-rigid object segmentation and shape reconstruction and its application to various applications such as free-view rendering and virtual reality.

With the increasing global popularity of self-driving cars, there is an immediate need for challenging real-world datasets for benchmarking and training various computer vision tasks such as 3D object detection. Existing datasets either represent simple scenarios or provide only day-time data. In this paper, we introduce a new challenging A*3D dataset which consists of RGB images and LiDAR data with a significant diversity of scene, time, and weather. The dataset consists of high-density images (≈ 10 times more than the pioneering KITTI dataset), heavy occlusions, a large number of nighttime frames (≈ 3 times the nuScenes dataset), addressing the gaps in the existing datasets to push the boundaries of tasks in autonomous driving research to more challenging highly diverse environments. The dataset contains 39K frames, 7 classes, and 230K 3D object annotations. An extensive 3D object detection benchmark evaluation on the A*3D dataset for various attributes such as high density, day-time/night-time, gives interesting insights into the advantages and limitations of training and testing 3D object detection in real-world setting.

Abstract Simultaneous semantically coherent object-based long-term 4D scene flow estimation, co-segmentation and reconstruction is proposed exploiting the coherence in semantic class labels both spatially, between views at a single time instant, and temporally, between widely spaced time instants of dynamic objects with similar shape and appearance. In this paper we propose a framework for spatially and temporally coherent semantic 4D scene flow of general dynamic scenes from multiple view videos captured with a network of static or moving cameras. Semantic coherence results in improved 4D scene flow estimation, segmentation and reconstruction for complex dynamic scenes. Semantic tracklets are introduced to robustly initialize the scene flow in the joint estimation and enforce temporal coherence in 4D flow, semantic labelling and reconstruction between widely spaced instances of dynamic objects. Tracklets of dynamic objects enable unsupervised learning of long-term flow, appearance and shape priors that are exploited in semantically coherent 4D scene flow estimation, co-segmentation and reconstruction. Comprehensive performance evaluation against state-of-the-art techniques on challenging indoor and outdoor sequences with hand-held moving cameras shows improved accuracy in 4D scene flow, segmentation, temporally coherent semantic labelling, and reconstruction of dynamic scenes.

Light fields are becoming an increasingly popular method of digital content production for visual effects and virtual/augmented reality as they capture a view dependent representation enabling photo realistic rendering over a range of viewpoints. Light field video is generally captured using arrays of cameras resulting in tens to hundreds of images of a scene at each time instance. An open problem is how to efficiently represent the data preserving the view-dependent detail of the surface in such a way that is compact to store and efficient to render. In this paper we show that constructing an Eigen texture basis representation from the light field using an approximate 3D surface reconstruction as a geometric proxy provides a compact representation that maintains view-dependent realism. We demonstrate that the proposed method is able to reduce storage requirements by > 95% while maintaining the visual quality of the captured data. An efficient view-dependent rendering technique is also proposed which is performed in eigen space allowing smooth continuous viewpoint interpolation through the light field.

We introduce the first approach to solve the challenging problem of unsupervised 4D visual scene understanding for complex dynamic scenes with multiple interacting people from multi-view video. Our approach simultaneously estimates a detailed model that includes a per-pixel semantically and temporally coherent reconstruction, together with instance-level segmentation exploiting photo-consistency, semantic and motion information. We further leverage recent advances in 3D pose estimation to constrain the joint semantic instance segmentation and 4D temporally coherent reconstruction. This enables per person semantic instance segmentation of multiple interacting people in complex dynamic scenes. Extensive evaluation of the joint visual scene understanding framework against state-of-the-art methods on challenging indoor and outdoor sequences demonstrates a significant (approx 40%) improvement in semantic segmentation, reconstruction and scene flow accuracy.

A common problem in wide-baseline matching is the sparse and non-uniform distribution of correspondences when using conventional detectors, such as SIFT, SURF, FAST, A-KAZE, and MSER. In this paper, we introduce a novel segmentation-based feature detector (SFD) that produces an increased number of accurate features for wide-baseline matching. A multi-scale SFD is proposed using bilateral image decomposition to produce a large number of scale-invariant features for wide-baseline reconstruction. All input images are over-segmented into regions using any existing segmentation technique, such as Watershed, Mean-shift, and simple linear iterative clustering. Feature points are then detected at the intersection of the boundaries of three or more regions. The detected feature points are local maxima of the image function. The key advantage of feature detection based on segmentation is that it does not require global threshold setting and can, therefore, detect features throughout the image. A comprehensive evaluation demonstrates that SFD gives an increased number of features that are accurately localized and matched between wide-baseline camera views; the number of features for a given matching error increases by a factor of 3-5 compared with SIFT; feature detection and matching performance are maintained with increasing baseline between views; multi-scale SFD improves matching performance at varying scales. Application of SFD to sparse multi-view wide-baseline reconstruction demonstrates a factor of 10 increases in the number of reconstructed points with improved scene coverage compared with SIFT/MSER/A-KAZE. Evaluation against ground-truth shows that SFD produces an increased number of wide-baseline matches with a reduced error.

Light-field video has recently been used in virtual and augmented reality applications to increase realism and immersion. However, existing light-field methods are generally limited to static scenes due to the requirement to acquire a dense scene representation. The large amount of data and the absence of methods to infer temporal coherence pose major challenges in storage, compression and editing compared to conventional video. In this paper, we propose the first method to extract a spatio-temporally coherent light-field video representation. A novel method to obtain Epipolar Plane Images (EPIs) from a spare lightfield camera array is proposed. EPIs are used to constrain scene flow estimation to obtain 4D temporally coherent representations of dynamic light-fields. Temporal coherence is achieved on a variety of light-field datasets. Evaluation of the proposed light-field scene flow against existing multi-view dense correspondence approaches demonstrates a significant improvement in accuracy of temporal coherence.

In this paper we propose a framework for spatially and temporally coherent semantic co-segmentation and reconstruction of complex dynamic scenes from multiple static or moving cameras. Semantic co-segmentation exploits the coherence in semantic class labels both spatially, between views at a single time instant, and temporally, between widely spaced time instants of dynamic objects with similar shape and appearance. We demonstrate that semantic coherence results in improved segmentation and reconstruction for complex scenes. A joint formulation is proposed for semantically coherent object-based co-segmentation and reconstruction of scenes by enforcing consistent semantic labelling between views and over time. Semantic tracklets are introduced to enforce temporal coherence in semantic labelling and reconstruction between widely spaced instances of dynamic objects. Tracklets of dynamic objects enable unsupervised learning of appearance and shape priors that are exploited in joint segmentation and reconstruction. Evaluation on challenging indoor and outdoor sequences with hand-held moving cameras shows improved accuracy in segmentation, temporally coherent semantic labelling and 3D reconstruction of dynamic scenes.

This paper presents an approach for reconstruction of 4D temporally coherent models of complex dynamic scenes. No prior knowledge is required of scene structure or camera calibration allowing reconstruction from multiple moving cameras. Sparse-to-dense temporal correspondence is integrated with joint multi-view segmentation and reconstruction to obtain a complete 4D representation of static and dynamic objects. Temporal coherence is exploited to overcome visual ambiguities resulting in improved reconstruction of complex scenes. Robust joint segmentation and reconstruction of dynamic objects is achieved by introducing a geodesic star convexity constraint. Comparative evaluation is performed on a variety of unstructured indoor and outdoor dynamic scenes with hand-held cameras and multiple people. This demonstrates reconstruction of complete temporally coherent 4D scene models with improved nonrigid object segmentation and shape reconstruction.

A common problem in wide-baseline stereo is the sparse and non-uniform distribution of correspondences when using conventional detectors such as SIFT, SURF, FAST and MSER. In this paper we introduce a novel segmentation based feature detector SFD that produces an increased number of 'good' features for accurate wide-baseline reconstruction. Each image is segmented into regions by over-segmentation and feature points are detected at the intersection of the boundaries for three or more regions. Segmentation-based feature detection locates features at local maxima giving a relatively large number of feature points which are consistently detected across wide-baseline views and accurately localised. A comprehensive comparative performance evaluation with previous feature detection approaches demonstrates that: SFD produces a large number of features with increased scene coverage, detected features are consistent across wide-baseline views for images of a variety of indoor and outdoor scenes, and the number of wide-baseline matches is increased by an order of magnitude compared to alternative detector-descriptor combinations. Sparse scene reconstruction from multiple wide-baseline stereo views using the SFD feature detector demonstrates at least a factor six increase in the number of reconstructed points with reduced error distribution compared to SIFT when evaluated against ground-truth and similar computational cost to SURF/FAST.

This paper introduces a general approach to dynamic scene reconstruction from multiple moving cameras without prior knowledge or limiting constraints on the scene structure, appearance, or illumination. Existing techniques or dynamic scene reconstruction from multiple wide-baseline camera views primarily focus on accurate reconstruction in controlled environments, where the cameras are fixed and calibrated and background is known. These approaches are not robust for general dynamic scenes captured with sparse moving cameras. Previous approaches for outdoor dynamic scene reconstruction assume prior knowledge of the static background appearance and structure. The primary contributions of this paper are twofold: an automatic method for initial coarse dynamic scene segmentation and reconstruction without prior knowledge of background appearance or structure, and a general robust approach for joint segmentation refinement and dense reconstruction of dynamic scenes from multiple wide-baseline static or moving cameras. Evaluation is performed on a variety of indoor and outdoor scenes with cluttered backgrounds and multiple dynamic non-rigid objects such as people. Comparison with state-of-the-art approaches demonstrates improved accuracy in both multiple view segmentation and dense reconstruction. The proposed approach also eliminates the requirement for prior knowledge of scene structure and appearance.