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
- Shaogang GongSuggested from co-authorship
- Loong‐Fah CheongSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works374 from public data
- Learning to Compare: Relation Network for Few-Shot Learning4,817
A conceptually simple, flexible, and general framework for few-shot learning, where a classifier must learn to recognise new classes given only few examples from each, which is easily extended to zero- shot learning.
- Rethinking Semantic Segmentation from a Sequence-to-Sequence Perspective with Transformers3,852
This paper deploys a pure transformer to encode an image as a sequence of patches, termed SEgmentation TRansformer (SETR), and shows that SETR achieves new state of the art on ADE20K, Pascal Context, and competitive results on Cityscapes.
- Deep Learning for Person Re-Identification: A Survey and Outlook2,304
A powerful AGW baseline is designed, achieving state-of-the-art or at least comparable performance on twelve datasets for four different Re-ID tasks, and a new evaluation metric (mINP) is introduced, indicating the cost for finding all the correct matches, which provides an additional criteria to evaluate the Re- ID system for real applications.
- Deep Mutual Learning2,054
Surprisingly, it is revealed that no prior powerful teacher network is necessary - mutual learning of a collection of simple student networks works, and moreover outperforms distillation from a more powerful yet static teacher.
- Domain Generalization: A Survey1,753
For the first time, a comprehensive literature review in DG is provided to summarize the developments over the past decade and conduct a thorough review into existing methods and theories.
- Omni-Scale Feature Learning for Person Re-Identification1,081
A novel deep ReID CNN is designed, termed Omni-Scale Network (OSNet), for omni-scale feature learning by designing a residual block composed of multiple convolutional feature streams, each detecting features at a certain scale.
- Semantic Autoencoder for Zero-Shot Learning911
This work presents a novel solution to ZSL based on learning a Semantic AutoEncoder (SAE), which outperforms significantly the existing ZSL models with the additional benefit of lower computational cost and beats the state-of-the-art when the SAE is applied to supervised clustering problem.
- Learning a Deep Embedding Model for Zero-Shot Learning791
This paper proposes to use the visual space as the embedding space instead of embedding into a semantic space or an intermediate space, and argues that in this space, the subsequent nearest neighbour search would suffer much less from the hubness problem and thus become more effective.
- Person Re-Identification by Support Vector Ranking772
This work converts the person re-identification problem from an absolute scoring p roblem to a relative ranking problem and develops an novel Ensemble RankSVM to overcome the scalability limitation problem suffered by existing SVM-based ranking methods.
- 747
- Learning a Discriminative Null Space for Person Re-identification669
This work proposes to overcome the SSS problem in re-id distance metric learning by matching people in a discriminative null space of the training data, which has a fixed dimension, a closed-form solution and is very efficient to compute.
- Learning to Generate Novel Domains for Domain Generalization566
This paper employs a data generator to synthesize data from pseudo-novel domains to augment the source domains, and outperforms current state-of-the-art DG methods on four benchmark datasets.
- Multi-level Factorisation Net for Person Re-identification564
Multi-Level Factorisation Net (MLFN), a novel network architecture that factorises the visual appearance of a person into latent discriminative factors at multiple semantic levels without manual annotation, achieves state-of-the-art results on three Re-ID datasets, as well as compelling results on the general object categorisation CIFAR-100 dataset.
- Transductive Multi-View Zero-Shot Learning548
A novel heterogeneous multi-view hypergraph label propagation method is formulated for zero-shot learning in the transductive embedding space that rectifies the projection shift between the auxiliary and target domains, exploits the complementarity of multiple semantic representations, and significantly outperforms existing methods for both zero- shot and N-shot recognition.
- Deep Reinforcement Learning for Unsupervised Video Summarization With Diversity-Representativeness Reward520
This paper forms video summarization as a sequential decision-making process and develops a deep summarization network (DSN) to summarize videos, which is comparable to or even superior than most of published supervised approaches.
- Deep Domain-Adversarial Image Generation for Domain Generalisation511
This paper proposes a novel DG approach based on Deep Domain-Adversarial Image Generation based on augmenting the source training data with the generated unseen domain data to make the label classifier more robust to unknown domain changes.
- Associating Groups of People510
A novel people group representation and a group matching algorithm are proposed that addresses changes in the relative positions of people in a group and the latter deals with variations in illumination and viewpoint across camera views.
- 488
- Pose-Normalized Image Generation for Person Re-identification483
This work proposes a novel deep person image generation model for synthesizing realistic person images conditional on the pose based on a generative adversarial network designed specifically for pose normalization in re-id, thus termed pose-normalization GAN (PN-GAN).
- 479
- 446
- 442
- 423
- 349
- Towards artificial general intelligence via a multimodal foundation model348
A multimodal foundation model is proposed that demonstrates the cross-domain learning and adaptation for broad range of downstream cognitive tasks and makes a transformative stride towards AGI.
- Sketch-a-Net: A Deep Neural Network that Beats Humans345
It is shown that state-of-the-art deep networks specifically engineered for photos of natural objects fail to perform well on sketch recognition, regardless whether they are trained using photos or sketches.
- Gait recognition without subject cooperation333
It is argued that selecting the most relevant gait features that are invariant to changes in gait covariate conditions is the key to develop a gait recognition system that works without subject cooperation and an Adaptive Component and Discriminant Analysis is formulated which seamlessly integrates the feature selection method with subspace analysis for robust recognition.
- Learning Generalisable Omni-Scale Representations for Person Re-Identification318
Novel CNN architectures to address the challenges of effective person re-identification and generalisable feature learning are developed, including a re-ID CNN termed omni-scale network (OSNet) to learn features that not only capture different spatial scales but also encapsulate a synergistic combination of multiple scales.
- Deep Transfer Learning for Person Re-Identification316
A two-stepped fine-tuning strategy with proxy classifier learning is developed to transfer knowledge from auxiliary datasets to address the training data sparsity problem from the supervised and unsupervised settings.
- 314
- Multi-scale Deep Learning Architectures for Person Re-identification311
A novel multi-scale deep learning model is proposed that is able to learn deep discriminative feature representations at different scales and automatically determine the most suitable scales for matching.
- 308
- 307
- Gait recognition using gait entropy image300
A novel gait representation termed as Gait Entropy Image (GEnI) is proposed, based on computing entropy, which encodes in a single image the randomness of pixel values in the silhouette images over a complete gait cycle that captures mostly motion information and is robust to covariate condition changes that affect appearance.
- 275
- Beyond Tracking: Modelling Activity and Understanding Behaviour271
A Dynamically Multi-Linked Hidden Markov Model (DML-HMM) is developed based on the discovery of salient dynamic interlinks among multiple temporal processes corresponding to multiple event classes resulting in its topology being intrinsically determined by the underlying causality and temporal order among events.
- Incremental Few-Shot Object Detection269
OpeN-ended Centre nEt (ONCE), a detector designed for incrementally learning to detect novel class objects with few examples, is proposed by an elegant adaptation of the efficient CentreNet detector to the few-shot learning scenario, and meta-learning a class-wise code generator model for registering novel classes.
- 269
- 252
- Transductive Multi-view Embedding for Zero-Shot Recognition and Annotation251
This paper proposes a novel framework, transductive multi-view embedding, that rectifies the projection shift between the auxiliary and target domains, exploits the complementarity of multiple semantic representations, achieves state-of-the-art recognition results on image and video benchmark datasets, and enables novel cross-view annotation tasks.
- Long-Term Cloth-Changing Person Re-identification248
This work introduces a shape embedding module as well as a cloth-elimination shape-distillation module aiming to eliminate the now unreliable clothing appearance features and focus on the body shape information.
- Sketch-a-Net that Beats Humans246
A multi-scale multi-channel deep neural network framework that yields sketch recognition performance surpassing that of humans, and not only delivers the best performance on the largest human sketch dataset to date, but also is small in size making efficient training possible using just CPUs.
- 245
- Simpler is Better: Few-shot Semantic Segmentation with Classifier Weight Transformer243
This work proposes to simplify the meta-learning task by focusing solely on the simplest component – the classifier, whilst leaving the en-coder and decoder to pre-training, and introduces a Classifier Weight Transformer (CWT) designed to dynamically adapt the support-set trained classifier’s weights to each query image in an inductive way.
- 240
- Crowd Counting and Profiling: Methodology and Evaluation226
This study describes and compares the state-of-the-art methods for video imagery based crowd counting, and provides a systematic evaluation of different methods using the same protocol.
- MixStyle Neural Networks for Domain Generalization and Adaptation218
MixStyle is a simple plug-and-play, parameter-free module that can improve domain generalization performance without the need to collect more data or increase model capacity, and can fit a variety of learning paradigms including supervised domain generalization, semi-supervised domain generalization, and unsupervised domain adaptation.
- Transferring a semantic representation for person re-identification and search214
This paper presents a new semantic attribute learning approach for person re-identification and search that is trained on existing fashion photography datasets and adapted to provide a powerful semantic description of surveillance person detections, without requiring any surveillance domain supervision.
- 213
- FLATTEN: optical FLow-guided ATTENtion for consistent text-to-video editing209
This paper introduces optical flow into the attention module in the diffusion model's U-Net to address the inconsistency issue for text-to-video editing and enforces the patches on the same flow path across different frames to attend to each other in the attention Module, thus improving the visual consistency in the edited videos.
- Person Re-identification199
This text explores the design rationale and implementation considerations of building a practical re-identification system, and introduces examples of robust feature representations and reviews salient feature weighting and selection mechanisms.
- 189
- Recent Advances in Zero-Shot Recognition: Toward Data-Efficient Understanding of Visual Content186
This article provides a comprehensive review of existing zero-shot recognition techniques covering various aspects ranging from representations of models, data sets, and evaluation settings and highlights the limitations of existing approaches.
- Stochastic Classifiers for Unsupervised Domain Adaptation185
This paper introduces a novel method called STochastic clAssifieRs (STAR) for addressing the problem of misaligned local regions between source and target domain, which finds that using more classifiers leads to better performance, but also introduces more model parameters, therefore risking overfitting.
- Spectral clustering with eigenvector selection185
A novel spectral clustering algorithm is proposed which differs from previous approaches in that only informative/relevant eigenvectors are employed for determining the number of clusters and performing clustering and the key element of the proposed algorithm is a simple but effective relevance learning method.
- Multi-camera Matching using Bi-Directional Cumulative Brightness Transfer Functions176
A Cumulative Brightness Transfer Function (CBTF) is proposed for mapping colour between cameras located at different physical sites, which makes better use of the available colour information from a very sparse training set and a bi-directional mapping approach is developed to obtain a more accurate similarity measure between a pair of candidate objects.
- 175
- Style-Based Global Appearance Flow for Virtual Try-On174
A novel global appearance flow estimation model is proposed in this work which enables the use of a global style vector to encode a whole-image context to cope with the aforementioned challenges.
- 168
- 168
- 165
- CLIP for All Things Zero-Shot Sketch-Based Image Retrieval, Fine-Grained or Not163
The proposed CLIP and prompt learning paradigm carries great promise in tackling other sketch-related tasks (not limited to ZS-SBIR) where data scarcity remains a great challenge.
- Deep Learning for Free-Hand Sketch: A Survey and A Toolbox160
A comprehensive survey of the deep learning techniques oriented at free-hand sketch data, and the applications that they enable.
- 157
- 148
- 145
- Text-Based Person Search with Limited Data144
This paper introduces a cross-modal momentum contrastive learning framework to enrich the training data for a given mini-batch of text-based person search and proposes to transfer knowledge learned from existing coarse-grained large-scale datasets containing image-text pairs from drastically different problem domains to compensate for the lack of TBPS training data.
- 141
- Learning to Sketch with Shortcut Cycle Consistency137
A novel approach for translating an object photo to a sketch, mimicking the human sketching process, and shows that the synthetic sketches can be used to train a better fine-grained sketch-based image retrieval model, effectively alleviating the problem of sketch data scarcity.
- SketchMate: Deep Hashing for Million-Scale Human Sketch Retrieval135
This work proposes a deep hashing framework for sketch retrieval that, for the first time, works on a multi-million scale human sketch dataset and shows that state-of-the-art hashing models specifically engineered for static images fail to perform well on temporal sketch data.
- Sketch Less for More: On-the-Fly Fine-Grained Sketch-Based Image Retrieval133
A reinforcement learning based cross-modal retrieval framework that directly optimizes rank of the ground-truth photo over a complete sketch drawing episode and introduces a novel reward scheme that circumvents the problems related to irrelevant sketch strokes, and thus provides us with a more consistent rank list during the retrieval.
- 129
- In Defence of Negative Mining for Annotating Weakly Labelled Data128
This work proposes a novel approach to annotating weakly labelled data, which performs image annotation by selecting exemplars that have never occurred before in the much larger, and strongly annotated, negative training set (maximising inter-class variance).
- Dictionary Learning with Iterative Laplacian Regularisation for Unsupervised Person Re-identification126
A new dictionary learning for sparse coding formulation with a graph Laplacian regularisation term whose value is set iteratively enables the exploitation of cross-view identity-discriminative information ignored by existing unsupervised Re-ID methods.
- Few-Shot Learning With Global Class Representations121
This paper proposes to tackle the challenging few-shot learning (FSL) problem by learning global class representations using both base and novel class training samples, and an effective sample synthesis strategy is developed to avoid overfitting.
- Person Re-Identification by Unsupervised $$\ell _1$$ Graph Learning120
This work proposes a novel unsupervised Re-ID approach which requires no labelled training data yet is able to capture discriminative information for cross-view identity matching, and is based on a new graph regularised dictionary learning algorithm.
- Unsupervised Learning of Generative Topic Saliency for Person Re-identification119
This document is distributed unchanged freely in print or electronic forms because the authors believe it is in the public interest to do so.
- Attribute Learning for Understanding Unstructured Social Activity119
This paper contributes an unstructured social activity attribute (USAA) dataset, introduces the concept of semi-latent attribute space, and proposes a novel model for learning the latent attributes which alleviate the dependence of existing models on exact and exhaustive manual specification of the attribute-space.
- Cross View Gait Recognition Using Correlation Strength119
A novel approach to cross-view gait recognition with the view angle of a probe gait sequence unknown is developed, which can cope with feature mis-match across view and is more robust against feature noise.
- 118
- GenTron: Diffusion Transformers for Image and Video Generation117
The initial step was to adapt Diffusion Transformers from class to text conditioning, a process involving thorough empirical exploration of the conditioning mechanism, and GenTron was introduced, a family of Generative models employing Transformer-based diffusion to address the gap in visual quality.
- Learning from Weak and Noisy Labels for Semantic Segmentation116
This work casts the WSSS problem into a label noise reduction problem, and develops an efficient learning algorithm by introducing an intermediate labelling variable to directly and explicitly detect noisy labels.
- Video Behaviour Mining Using a Dynamic Topic Model116
The MCTM builds on existing dynamic Bayesian network models and Bayesian topic models, and overcomes their drawbacks on sensitivity, robustness and efficiency, and profiles complex dynamic scenes by robustly clustering visual events into activities and these activities into global behaviours with temporal dynamics.
- Human-in-the-Loop Person Re-identification115
A novel human-in-the-loop re-id model based on Human Verification Incremental Learning (HVIL) is formulated which does not require any pre-labelled training data to learn a model, therefore readily scalable to new camera pairs, and is formulated to combine a series of incrementally learned HVIL models into a single ensemble model to be used when human feedback becomes unavailable.
- StyleMeUp: Towards Style-Agnostic Sketch-Based Image Retrieval112
This model can not only disentangle the cross-modal shared semantic content for SBIR, but can adapt the disentanglement to any unseen user style as well, making the SBIr model truly style-agnostic.
- Semantic Regularisation for Recurrent Image Annotation111
This work proposes to use a semantically regularised embedding layer as the interface between the CNN and RNN, allowing each to be more effectively trained and jointly training much more efficient.
- Learning Deep Sketch Abstraction109
This work proposes the first stroke-level sketch abstraction model based on the insight of sketch abstraction as a process of trading off between the recognizability of a sketch and the number of strokes used to draw it and shows that the model can be used for various sketch analysis tasks.
- Incremental and adaptive abnormal behaviour detection109
The experimental results show that the incremental and adaptive behaviour modelling approach is superior to a conventional batch-mode one in terms of both performance on abnormality detection and computational efficiency.
- 105
- 103
- 102
- Domain-Adaptive Few-Shot Learning101
This work proposes a novel domain-adversarial prototypical network (DAPN) model, designed to address a specific challenge in DA-FSL, which dictates that the target domain per class distribution must be different from that of any source domain class, meaning aligning the distributions across domains may harm the FSL performance.
- 101
- 101
- 99
- 96
- 96
- 94
- SketchyScene: Richly-Annotated Scene Sketches93
The potential impact of SketchyScene is demonstrated by training new computational models for semantic segmentation of scene sketches and showing how the new dataset enables several applications including image retrieval, sketch colorization, editing, and captioning, etc.
- 93
- Interestingness Prediction by Robust Learning to Rank91
A more principled way to identify annotation outliers is proposed by formulating theinterestingness prediction task as a unified robust learning to rank problem, tackling both the outlier detection and interestingness prediction tasks jointly.
- Cross-modal subspace learning for fine-grained sketch-based image retrieval90
Through thorough examination of the experimental results, it is demonstrated that the subspace learning can effectively model the sketch-photo domain-gap, and a few key insights to drive future research are drawn.
- Fusing appearance and distribution information of interest points for action recognition90
This paper proposes a novel action representation method which differs significantly from the existing interest point based representation in that only the global distribution information of interest points is exploited and holistic features from clouds ofinterest points accumulated over multiple temporal scales are extracted.
- Head Pose Classification in Crowded Scenes88
A new head pose descriptor is formulated using similarity distance maps by indexing each pixel of a head image to the mean appearance templates of head images at different poses, which are then used to train a multi-class Support Vector Machine for pose classification.
- Gait Representation Using Flow Fields88
This paper proposes a novel gait representation based on optical flow fields computed from normalized and centred person images over a complete gait cycle that is not only more discriminative, but also less sensitive to changes in various covariate conditions including clothing, carrying, shoe, and speed.
- 88
- FAME-ViL: Multi-Tasking Vision-Language Model for Heterogeneous Fashion Tasks86
This work proposes a novel FAshion-focused Multi-task Efficient learning method for Vision-and-Language tasks (FAME-ViL), which applies a single model for multiple heterogeneous fashion tasks, therefore being much more parameter-efficient.
- Multi-Modal Multi-Scale Deep Learning for Large-Scale Image Annotation86
A novel two-branch deep neural network architecture is proposed, which comprises a very deep main network branch and a companion feature fusion network branch designed for fusing the multi-scale features computed from the main branch.
- 85
- Scene Segmentation for Behaviour Correlation85
A novel framework for detecting abnormal pedestrian and vehicle behaviour by modelling cross-correlation among different co-occurring objects both locally and globally in a given scene is presented.
- 80
- Detecting and discriminating behavioural anomalies80
The proposed framework outperforms existing methods in detecting durational anomalies in complex behaviours and subtle anomalies that are difficult to detect when objects are viewed in isolation, allowing for a more effective detection of subtle anomalies in surveillance videos.
- Single-cell multi-stage spatial evolutional map of esophageal carcinogenesis78
A spatial transcriptomic map at single-cell resolution is presented, encompassing 127 multi-stage fields of view from 43 patients, to chart the evolutionary trajectories of human esophageal squamous cell carcinoma (ESCC).
- Hybrid Graph Neural Networks for Few-Shot Learning78
A novel hybrid GNN (HGNN) model consisting of two GNNs, an instance GNN and a prototype GNN designed to deal with a fundamental yet often neglected challenge in FSL, that is, with only a handful of shots per class, any few-shot classifier would be sensitive to badly sampled shots.
- Gait Recognition by Ranking78
This work proposes a novel approach which casts gait recognition as a bipartite ranking problem and leverages training samples from different classes/people and even from different datasets, which makes this approach suitable for recognition under a genuine uncooperative setting and robust against any covariate types.
- Zero and Few Shot Learning With Semantic Feature Synthesis and Competitive Learning77
A novel semantic data synthesis strategy is proposed, by which semantic class prototypes are used to simply perturb seen class data for generating unseen class ones by exploiting semantic (class prototype guided) feature synthesis and competitive BPL.
- BézierSketch: A Generative Model for Scalable Vector Sketches77
This paper introduces a novel inverse graphics approach to stroke embedding that trains an encoder to embed each stroke to its best fit Bezier curve, which enables to treat sketches as short sequences of paramaterized strokes and thus train a recurrent sketch generator with greater capacity for longer sketches, while producing scalable high-resolution results.
- Face-Focused Cross-Stream Network for Deception Detection in Videos77
A novel face-focused cross-stream network (FFCSN) is proposed for face-body multimodal learning and its robust training strategy are shown to be generally applicable to other human-centric video analysis tasks such as emotion recognition from user-generated videos.
- 76
- Picture that Sketch: Photorealistic Image Generation from Abstract Sketches74
This paper turns an abstract, deformed, ordinary sketch from untrained amateurs like you and me into a photorealistic image, and proposes an autoregressive sketch mapper trained on sketch-photo pairs that maps a sketch to the StyleGAN latent space.
- Sketch3T: Test-Time Training for Zero-Shot SBIR73
This paper extends ZS-SBIR to include a test-time training paradigm that can adapt using just one sketch, and designs a novel meta-learning based training paradigm to learn a separation between model updates incurred by this auxiliary task from those off the primary objective of discriminative learning.
- 73
- Security and Surveillance72
A significant application of video analysis and understanding is intelligent surveillance, which aims to interpret automatically human activity and detect unusual events that could pose a threat to public security and safety.
- More Photos are All You Need: Semi-Supervised Learning for Fine-Grained Sketch Based Image Retrieval71
This paper introduces a novel semi-supervised framework for cross-modal retrieval that can additionally leverage large-scale unlabelled photos to account for data scarcity, and treats generation and retrieval as two conjugate problems, where a joint learning procedure is devised for each module to mutually benefit from each other.
- Vectorization and Rasterization: Self-Supervised Learning for Sketch and Handwriting69
This paper proposes two novel cross-modal translation pre-text tasks for self-supervised feature learning: Vectorization and Rasterization and shows that the learned encoder modules benefit both raster-based and vector-based downstream approaches to analysing hand-drawn data.
- 69
- Bayesian Joint Topic Modelling for Weakly Supervised Object Localisation69
All object classes and image backgrounds are modelled jointly together in a single generative model so that "explaining away" inference can resolve ambiguity and lead to better learning and localisation.
- Discriminative Topics Modelling for Action Feature Selection and Recognition68
A novel framework for recognising realistic human actions in unconstrained environments based on computing a rich set of descriptors from key point trajectories is presented and an adaptive feature fusion method to combine different local motion descriptors for improving model robustness against feature noise and background clutters is developed.
- HeadSculpt: Crafting 3D Head Avatars with Text67
This work introduces a versatile coarse-to-fine pipeline dubbed HeadSculpt for crafting (i.e., generating and editing) 3D head avatars from textual prompts, and proposes a novel identity-aware editing score distillation strategy to optimize a textured mesh with a high-resolution differentiable rendering technique.
- Uncooperative gait recognition by learning to rank67
This work proposes a novel approach which casts gait recognition as a bipartite ranking problem and leverages training samples from different people and even from different datasets, which makes this approach suitable under a genuine uncooperative setting and robust against changes in any covariate conditions.
- Boundary-sensitive Pre-training for Temporal Localization in Videos66
This paper proposes to synthesize temporal boundaries in existing video action classification datasets and introduces a novel boundary-sensitive pretext (BSP) task, which is superior and complementary to the existing action classification-based pre-training counterpart and achieves new state-of-the-art performance on several temporal localization tasks.
- Proposal-Free Temporal Action Detection via Global Segmentation Mask Learning65
The TAGS model differs significantly from the conventional proposal-based methods by focusing on global temporal representation learning to directly detect local start and end points of action instances without proposals, and needs a much simpler model architecture with lower computational cost.
- 65
- Background Subtraction with Dirichlet Processes65
A new method is presented that uses a Dirichlet process Gaussian mixture model to estimate a per-pixel background distribution, which is followed by probabilistic regularisation.
- Context-Aware Layout to Image Generation with Enhanced Object Appearance64
A context-aware feature transformation module is introduced in the generator to ensure that the generated feature encoding of either object or stuff is aware of other coexisting objects/stuff in the scene.
- Few-shot Action Recognition with Prototype-centered Attentive Learning64
A prototype-centered contrastive learning loss is introduced to complement the conventional query-centered learning objective, in order to make full use of the limited training samples in each episode of few-shot action recognition.
- Cross-Modal Hierarchical Modelling for Fine-Grained Sketch Based Image Retrieval62
A novel network is designed that is capable of cultivating sketch-specific hierarchies and exploiting them to match sketch with photo at corresponding hierarchical levels, and enriched using cross-modal co-attention and hierarchical node fusion at every level to form a better embedding space to conduct retrieval.
- 60
- Sketching without Worrying: Noise-Tolerant Sketch-Based Image Retrieval59
This paper proposes an auxiliary module for existing retrieval models that predominantly lets the users sketch without having to worry, and designs a stroke subset selector that detects noisy strokes, leaving only those which make a positive contribution towards successful retrieval.
- Disjoint Label Space Transfer Learning with Common Factorised Space59
A unified approach is presented to transfer learning that addresses several source and target domain label-space and annotation assumptions with a single model that outperforms alternatives in both unsupervised and semi-supervised settings.
- Weakly Supervised Action Detection59
For the first time, a weakly supervised action detection method is proposed which only requires binary labels of the videos indicating the presence of the action of interest, and that the proposed MIL algorithm significantly outperforms the existing ones.
- Robust Subjective Visual Property Prediction from Crowdsourced Pairwise Labels58
A more principled way to identify annotation outliers is proposed by formulating the subjective visual property prediction task as a unified robust learning toRank problem, tackling both the outlier detection and learning to rank jointly.
- Aberrant epithelial cell interaction promotes esophageal squamous-cell carcinoma development and progression56
It is demonstrated that the aberrant epithelial cell interaction via EFNB1-EPHB4 triggers EMT and cell cycle mediated by downstream SRC/ERK/AKT signaling, and the aberrant EFNB1-EPHB4 interaction is caused by overexpressed ΔNP63 due to TP53 mutation, the culprit in human ESCC tumorigenesis.
- Fine-Grained Instance-Level Sketch-Based Image Retrieval56
A new large-scale FG-SBIR database is introduced which is carefully designed to reflect the real-world application scenarios and a deep cross-domain matching model is formulated to solve the intrinsic drawing style variability, large domain gap issues, and capture instance-level discriminative features.
- Transductive Multi-label Zero-shot Learning56
This paper proposes a multi-output deep regression model to project an image into a semantic word space, which explicitly exploits the correlations in the intermediate semantic layer of word vectors, and a novel zero-shot learning algorithm for multi-label data that exploits the unique compositionality property of semantic word vector representations.
- 54
- iVQA: Inverse Visual Question Answering54
This work poses question generation as a multi-modal dynamic inference process and proposes an iVQA model that can gradually adjust its focus of attention guided by both a partially generated question and the answer.
- Dynamic Instance Domain Adaptation53
A dynamic neural network with adaptive convolutional kernels is developed to generate instance-adaptive residuals to adapt domain-agnostic deep features to each individual instance in a dynamic instance domain adaptation (DIDA).
- Universal Sketch Perceptual Grouping53
A universal sketch grouper that can be applied to sketches of any category in any domain to group constituent strokes/segments into semantically meaningful object parts is developed and significantly outperforms the state-of-the-art groupers.
- Learning Behavioural Context52
A novel framework for automatic discovering and learning of behavioural context for video-based complex behaviour recognition and anomaly detection and a method for detecting subtle behavioural anomalies against the learned context are developed.
- MarDini: Masked Autoregressive Diffusion for Video Generation at Scale51
The new MarDini family of video diffusion models that integrate the advantages of masked auto-regression (MAR) into a unified diffusion model (DM) framework sets a new state-of-the-art for video interpolation and efficiently generates videos on par with those of much more expensive advanced image-to-video models.
- Weakly-Supervised Image Annotation and Segmentation with Objects and Attributes51
This work proposes a novel Weakly Supervised Markov Random Field Stacked Indian Buffet Process (WS-MRF-SIBP) that models objects and attributes as latent factors and explicitly captures their correlations within and across superpixels.
- 51
- Multi-camera Matching under Illumination Change Over Time51
This paper proposes a novel method for multi-camera object association based on adapting a learned inter- camera illumination mapping function to new illumination conditions over time without the need for a manual training stage using new fore- ground objects.
- It's All About Your Sketch: Democratising Sketch Control in Diffusion Models50
This paper unravels the potential of sketches for diffusion models, addressing the deceptive promise of direct sketch control in generative AI, and introduces an abstraction-aware framework, using a sketch adapter, adaptive time-step sampling, and discriminative guidance from a pre-trained fine-grained sketch-based image retrieval model, working synergistically to reinforce fine-grained sketch-photo association.
- 50
- Global Abnormal Behaviour Detection Using a Network of CCTV Cameras49
A novel framework is proposed, which consists of object tracking across camera views, global behaviour modelling based on unsupervised learning, and probabilistic abnormality inference, and the effectiveness of the framework is demonstrated with experiments on real-world surveillance video data.
- Joint Learning of Semantic and Latent Attributes48
It is argued that learning the latent attributes jointly with user-defined semantic attributes not only leads to better representation for object recognition but also helps with semantic attribute prediction.
- 47
- Expression recognition using fuzzy spatio-temporal modeling47
A novel and effective approach for facial expression recognition that analyzes a sequence of images (displaying one expression) instead of just one image (which captures the snapshot of an emotion) is proposed.
- Autonomous Visual Events Detection and Classification without Explicit Object-Centred Segmentation and Tracking47
A new approach for modelling such temporal events based on the local intensity temporal history of pixels is proposed, which provides a computationally very effective temporal measure for detecting autonomous events.
- What Can Human Sketches Do for Object Detection?46
A sketch-enabled object detection framework that detects based on what you sketch, and outperforms both supervised (SOD) and weakly-supervised object detectors (WSOD) on zero-shot setups.
- Deep Multi-task Attribute-driven Ranking for Fine-grained Sketch-based Image Retrieval46
A fine-grained SBIR model that exploits semantic attributes and deep feature learning in a complementary way is proposed, and the introduction of semantic attribute learning in the model allows for the elimination of the otherwise prohibitive cost of human annotations required for training a fine- grained deep ranking model.
- Stream-Based Active Unusual Event Detection46
Experimental results on busy public space videos show that with minimal human supervision, the approach outperforms existing supervised and unsupervised learning strategies in identifying unusual events.
- 46
- 45
- 44
- 44
- Pixelor44
The first competitive drawing agent Pixelor is presented that exhibits humanlevel performance at a Pictionary-like sketching game, and is better than the human players of the Quick, Draw! game, under both AI and human judging of early recognition.
- 41
- 41
- SceneTrilogy: On Human Scene-Sketch and its Complementarity with Photo and Text40
This paper extends scene understanding to include that of human sketch and creates a complete trilogy of scene representation from three diverse and complementary modalities – sketch, photo, and text by learning a flexible joint embedding that fully supports the “optionality” that this complementarity brings.
- Deep Ranking for Image Zero-Shot Multi-Label Classification40
This work investigates and formalizes a flexible framework consisting of two components, i.e., visual-semantic embedding and zero-shot multi-label prediction, and presents a deep regression model to project the visual features into the semantic space, which explicitly exploits the correlations in the intermediate semantic layer of word vectors and makes label prediction possible.
- Fine-Grained Instance-Level Sketch-Based Video Retrieval40
This work proposes a novel cross-modal retrieval problem of fine-grained instance-level sketch-based video retrieval (FG-SBVR), where a sketch sequence is used as a query to retrieve a specific target video instance.
- 40
- 40
- Cloud2Curve: Generation and Vectorization of Parametric Sketches39
Cloud2Curve is presented, a generative model for scalable high-resolution vector sketches that can be trained end-to-end using point-cloud data alone and is capable of deterministic vectorization which can map novel raster or waypoint based sketches to their corresponding high- resolution scalable Bézier equivalent.
- 39
- 39
- 38
- Learning Flow Fields in Attention for Controllable Person Image Generation37
This work proposes learning flow fields in attention (Leffa), which explicitly guides the target query to attend to the correct reference key in the attention layer during training via a regularization loss on top of the attention map within a diffusionbased baseline.
- DiffTAD: Temporal Action Detection with Proposal Denoising Diffusion37
A new formulation of temporal action detection with denoising diffusion with denoising diffusion, DiffTAD, which can yield action proposals accurately given an untrimmed long video, presents a generative modeling perspective, against previous discriminative learning manners.
- Modelling Multi-object Activity by Gaussian Processes36
Experimental results on a public traffic scene show that the GP models outperform DBNs in terms of anomaly sensitivity, noise robustness, and flexibility in modelling complex activity.
- Feature Selection for Gait Recognition without Subject Cooperation36
A novel gait recognition approach is proposed, which performs automatic feature selection on each pair gallery and probe gait sequences, and seamlessly integrates feature selection with an Adaptive Component and Discriminant Analysis (ACDA) for fast recognition.
- Uncertainty-Aware Source-Free Domain Adaptive Semantic Segmentation35
This paper proposes to use Bayesian Neural Network (BNN) to improve the target self- training by better estimating and exploiting pseudo-label uncertainty and introduces two novel self-training based components: Uncertainty-aware Online Teacher-Student Learning (UOTSL) and Uncertainly-aware FeatureMix (UFM).
- Prediction Calibration for Generalized Few-Shot Semantic Segmentation35
To make the proposed cross-attention module training tractable at the pixel level, this module is designed based on feature-score cross-covariance and episodically trained to be generalizable at inference time.
- How to Handle Sketch-Abstraction in Sketch-Based Image Retrieval?34
This paper proposes a novel abstraction-aware sketch-based image retrieval framework capable of handling sketch abstraction at varied levels, and proposes feature-level and retrieval granularity-level designs so that the system builds into its DNA the necessary means to interpret abstraction.
- Semantic Embedding for Sketch-Based 3D Shape Retrieval.34
This work argues that the large domain gap cannot be effectively bridged in a shared feature space and proposes a novel deep cross-domain semantic embedding model that drastically improves over the state-of-the-art alternatives.
- Action Detection in Crowd34
This paper proposes a novel greedy k nearest neighbour algorithm for automated annotation of positive training data, by which an action detector can be learned with only a single training sequence being annotated thus greatly alleviating the tedious and unreliable 3D manual annotation.
- Doodle It Yourself: Class Incremental Learning by Drawing a Few Sketches33
This paper addresses two key questions that bottleneck its ubiquitous application of FSCIL: can the model learn from diverse modalities other than just photo, and what if photos are not readily accessible (due to ethical and privacy constraints).
- Disentangled Lifespan Face Synthesis33
A novel LFS model is proposed to disentangle the key face characteristics including shape, texture and identity so that the unique shape and texture age transformations can be modeled effectively.
- Person Re-identification in Identity Regression Space33
This work proposes to construct an identity regression space (IRS) based on embedding different training person identities (classes) and formulate re-id as a regression problem solved by identity regression in the IRS, characterised by a closed-form solution with high learning efficiency and an inherent incremental learning capability with human-in theloop.
- 33
- Bayesian Joint Modelling for Object Localisation in Weakly Labelled Images33
A novel framework based on Bayesian joint topic modelling is proposed, which differs significantly from the existing ones in that all foreground object classes are modelled jointly in a single generative model that encodes multiple object co-existence so that “explaining away” inference can resolve ambiguity and lead to better learning and localisation.
- Doodle Your 3D: from Abstract Freehand Sketches to Precise 3D Shapes32
This paper introduces a novel part-level modelling and alignment framework that facilitates abstraction modelling and cross-modal correspondence that seamlessly extends to sketch modelling by establishing correspondence between CLIPasso edgemaps and projected 3D part regions.
- Exploiting Unlabelled Photos for Stronger Fine-Grained SBIR32
This paper advances the fine-grained sketch-based image retrieval (FG-SBIR) literature by putting forward a strong baseline that overshoots prior state-of-the-arts by ≈11 % and proposes a simple modification to the standard triplet loss, that explicitly enforces separation amongst photos/sketch instances.
- Partially Does It: Towards Scene-Level FG-SBIR with Partial Input31
This work advocates for a simple set-based approach using optimal transport (OT) to model cross-modal region associativity in a partially-aware fashion and improves upon OT to further account for holistic partialness by comparing intra- modal adjacency matrices.
- Inverse Visual Question Answering: A New Benchmark and VQA Diagnosis Tool31
This paper proposes a variational iVQA model that can generate diverse, grammatically correct and content correlated questions that match the given answer, and shows that iVZA is an interesting benchmark for visuo-linguistic understanding, and a more challenging alternative to VQA because an iV QA model needs to understand the image better to be successful.
- Im2Sketch: Sketch generation by unconflicted perceptual grouping31
This paper studies how multiple Gestalt rules can be encapsulated into a unified perceptual grouping framework for sketch generation and shows that by solving the problem of Gestalt confliction, more similar to human-made sketches can be generated.
- You'll Never Walk Alone: A Sketch and Text Duet for Fine-Grained Image Retrieval30
A novel compositionality framework is introduced, effectively combining sketches and text using pre-trained CLIP models, while eliminating the need for extensive fine-grained textual descriptions, providing solutions for various real-world scenarios.
- Self-Supervised Video Representation Learning with Constrained Spatiotemporal Jigsaw30
Constrained Spatiotemporal Jigsaw (CSJ) is proposed whereby the 3D jigsaws are formed in a constrained manner to ensure that large continuous spatiotem temporal cuboids exist, which provides sufficient cues for the model to reason about the continuity.
- Sketch-a-Classifier: Sketch-Based Photo Classifier Generation30
This paper investigates an alternative approach of synthesizing image classifiers: almost directly from a user's imagination, via freehand sketch, which can also be used to enhance the granularity of an existing photo classifier, or as a complement to name-based zero-shot learning.
- Learning to Generate and Edit Hairstyles30
A new large-scale hairstyle dataset called Hairstyle30k is contributed, which is composed of 30k images containing 64 different types of hairstyles and a novel GAN model termed Hairstyles GAN (H-GAN) which can be learned efficiently is proposed.
- Vision Transformers: From Semantic Segmentation to Dense Prediction29
This work explores the global context learning potentials of ViTs for dense visual prediction and forms a family of Hierarchical Local-Global (HLG) Transformers, characterized by local attention within windows and global-attention across windows in a pyramidal architecture.
- 29
- Adaptive Fine-Grained Sketch-Based Image Retrieval28
This paper introduces a novel model-agnostic meta-learning (MAML) based framework with several key modifications, and proposes a novel perspective -- instead of asking for a model that generalises, it advocates for one that quickly adapts, with just very few samples during testing (in a few-shot manner).
- AdarGCN: Adaptive Aggregation GCN for Few-Shot Learning28
This work defines a new FSL setting termed scarce-source few-shot learning (SSFSL), under which both the source and target classes have limited training samples, and proposes a graph convolutional network-based label denoising (LDN) method to remove irrelevant images.
- Understanding the Behavior of SFM Algorithms: A Geometric Approach28
A geometrically motivated motion error analysis capable of supporting investigation of global effect such as inherent ambiguities, in contrast with the usual statistical kinds of motion error analyses which can only deal with local effectsuch as noise perturbations.
- Text-to-Image Diffusion Models are Great Sketch-Photo Matchmakers27
This paper introduces a straightforward yet powerful strategy focused on two key aspects: selecting optimal feature layers and utilising visual and textual prompts to guide the model's feature extraction process, enabling it to generate more discriminative and contextually relevant cross-modal representations.
- Sketch2Saliency: Learning to Detect Salient Objects from Human Drawings27
A photo-to-sketch generation model that aims to generate sequential sketch coordinates corresponding to a given visual photo through a 2D attention mechanism to study how sketches can be used as a weak label to detect salient objects present in an image.
- Highly Efficient Regression for Scalable Person Re-Identification27
A Highly Efficient Regression (HER) model is formulated by embedding the Fisher's criterion to a ridge regression model for very fast re-id model learning with scalable memory/storage usage and shows that such a simple and fast model not only outperforms notably the state-of-the-art re-ID methods, but also is more scalable to large data with additional benefits to active learning.
- 27
- Softmax-Free Linear Transformers26
A family of SOftmax-Free Transformers (SOFT) is proposed, with a Gaussian kernel function adopted to replace the dot-product similarity, enabling a full self-attention matrix to be approximated under low-rank matrix decomposition.
- 26
- Where is my Wallet? Modeling Object Proposal Sets for Egocentric Visual Query Localization25
The proposed adaptations improve egocentric query detection, leading to a better visual query localization system in both 2D and 3D configurations, and a novel transformer-based module that allows for object-proposal set context to be considered while incorporating query information is proposed.
- Goal-Driven Sequential Data Abstraction25
This paper studies a general reinforcement learning based framework for learning to abstract sequential data in a goal-driven way and processes the input sequence holistically without being constrained by the original input order.
- 25
- 24
- Global Aggregation Then Local Distribution for Scene Parsing23
A novel local distribution module is designed which models the affinity map between global and local relationship for each pixel adaptively and can be modularized as an end-to-end trainable block and easily plugged into existing semantic segmentation networks, giving rise to the GALD networks.
- Transferrable Feature and Projection Learning with Class Hierarchy for Zero-Shot Learning23
A novel inductive ZSL model that leverages superclasses as the bridge between seen and unseen classes to narrow the domain gap is proposed and can be easily extended to a closely related task—few-shot learning (FSL).
- Few-Shot Temporal Action Localization with Query Adaptive Transformer22
A novel FS-TAL model is proposed which maximizes the knowledge transfer from training classes whilst enabling the model to be dynamically adapted to both the new class and each video of that class simultaneously.
- 22
- 3D VR Sketch Guided 3D Shape Prototyping and Exploration21
A 3D shape generation network that takes a 3D VR sketch as a condition, assuming that sketches are created by novices without art training and aim to reconstruct geometrically realistic 3D shapes of a given category is proposed.
- A Unifying Theory of Active Discovery and Learning21
This study shows with systematic theoretical analysis how the previously disparate tasks of active discovery and learning can be cleanly unified into a single problem, and is able for the first time to develop a unified query algorithm to directly optimise this problem.
- Finding Rare Classes: Adapting Generative and Discriminative Models in Active Learning21
A unified active learning model to jointly discover new categories and learn to classify them; and a classifier combination algorithm that switches generative and discriminative classifiers as learning progresses that demonstrates the superiority of this approach over existing methods.
- 21
- Wired Perspectives: Multi-View Wire Art Embraces Generative AI20
This work presents DreamWire, an AI system enabling everyone to craft MVWA easily, which synergises 3D Bézier curves, Prim's algorithm, and knowledge distillation from diffusion models or their variants (e.g., ControlNet).
- One Sketch for All: One-Shot Personalized Sketch Segmentation20
The first one-shot personalized sketch segmentation method, which aims to segment all sketches belonging to the same category provisioned with a single sketch with a given part annotation while preserving the parts semantics embedded in the exemplar, and being robust to input style and abstraction is presented.
- Active Learning using Dirichlet Processes for Rare Class Discovery and Classification20
A novel active learning approach is proposed, which automatically balances the competing goals of new class discovery and improving classification, and is free of tuneable parameters.
- 19
- ChiroDiff: Modelling chirographic data with Diffusion Models19
A powerful model-class namely Denoising Diffusion Probabilistic Models or DDPMs for chirographic data that learns to capture holistic concepts and therefore remains resilient to higher temporal sampling rate up to a good extent and it is shown that many important downstream utilities can be flexibly implemented using ChiroDiff.
- SketchDreamer: Interactive Text-Augmented Creative Sketch Ideation19
The objective is to empower non-professional users to create sketches and, through a series of optimisation processes, transform a narrative into a storyboard by expanding the text prompt while making minor adjustments to the sketch input.
- 19
- Deep Factorised Inverse-Sketching19
A novel unsupervised image style transfer model based on enforcing a cyclic embedding consistency constraint is proposed and evaluated both qualitatively and quantitatively to demonstrate its superiority over a number of state-of-the-art alternatives for style transfer and FG-SBIR.
- Active Rare Class Discovery and Classification Using Dirichlet Processes19
An active learning method for scenarios with unknown, rare classes, where the problems of classification and rare class discovery need to be tackled jointly, by assuming a non-parametric prior on the data.
- Model Selection for Unsupervised Learning of Visual Context19
This study suggests that for learning visual context using a mixture model, BICr is the most appropriate criterion given sparse data, while CL-AIC should be chosen given moderate or large data sample sizes.
- 19
- Move Anything with Layered Scene Diffusion18
This work proposes SceneDiffusion to optimize a layered scene representation during the diffusion sampling process, and finds that spatial disentanglement can be obtained by jointly denoising scene rende rings at different spatial layouts.
- SketchINR: A First Look into Sketches as Implicit Neural Representations18
SketchINR emulates the human ability to reproduce a sketch with varying abstraction in terms of number and complexity of strokes, and its compact high-fidelity representation will support future work in modelling long and complex sketches.
- SketchXAI: A First Look at Explainability for Human Sketches18
It is argued that sketch as a “human-centred” data form, represents a natural interface to study explainability, and a simple explainability-friendly sketch encoder is designed that accommodates the intrinsic properties of strokes: shape, location, and order.
- Towards 3D VR-Sketch to 3D Shape Retrieval18
This paper studies the use of 3D sketches as an input modality and advocates a VR-scenario where retrieval is conducted, and shows that volumetric point-based approaches exhibit superior performance on 3D sketch to 3D shape retrieval due to the sparse and abstract nature of3D VR-sketches.
- Weakly Supervised Learning of Objects, Attributes and Their Associations18
This paper models object-attribute associations from weakly labelled images, such as those widely available on media sharing sites (e.g. Flickr), where only image-level labels are given, without their locations and associations.
- Learning Rare Behaviours18
A novel topic model is proposed: Multi-Class Delta Latent Dirichlet Allocation which learns to model rare behaviours from a few weakly labelled videos as well as typical behaviours from uninteresting videos by collaboratively sharing features among all classes of footage.
- DI-MaskDINO: A Joint Object Detection and Instance Segmentation Model17
DI-MaskDINO model is proposed, the core idea of which is to improve the final performance by alleviating the detection-segmentation imbalance.
- 17
- 17
- 17
- 17
- Negative Frames Matter in Egocentric Visual Query 2D Localization16
A more efficient and effective solution to the VQ2D challenge, which brings the training loop from ~15 days to less than 24 hours, and achieves 0.17% spatial-temporal AP, which is 31% higher than the baseline.
- Fine-Grained VR Sketching: Dataset and Insights16
The first fine-grained dataset of 1,497 3D VR sketch and 3D shape pairs of a chair category with large shapes diversity is presented, and it is argued for the most convenient sketching scenario where the sketch consists of sparse lines and does not require any sketching skills, prior training or time-consuming accurate drawing.
- Group Association: Assisting Re-identification by Visual Context15
A novel people group representation and a group matching algorithm are proposed that addresses changes in the relative positions of people in a group and the latter uses the proposed group descriptors for measuring the similarity between two candidate images.
- Optimising dynamic graphical models for video content analysis14
It is demonstrated that CL-AIC is superior to the other scoring functions in building dynamic graphical models for solving two challenging problems in video content analysis: (1) content based surveillance video segmentation and (2) discovering causal/temporal relationships among visual events for group activity modelling.
- What Sketch Explainability Really Means for Downstream Tasks?13
A lightweight and portable explainability solution - a seamless plugin that integrates effortlessly with any pre-trained model, elim-inating the need for re-training is proposed, and the centrepiece to the solution is a stroke-level attribution map that takes different forms when linked with downstream tasks.
- Structure-Aware 3D VR Sketch to 3D Shape Retrieval13
This work proposes to use a triplet loss with an adaptive margin value driven by a ‘fitting gap’, which is the similarity of two shapes under structure-preserving deformations, and introduces a dataset of 202 VR sketches for 202 3D shapes drawn from memory rather than from observation.
- 13
- Learning Robust Graph Regularisation for Subspace Clustering13
This work identifies two critical limitations of the graph regularisation term employed in existing subspace clustering models and provides solutions for both of them and proposes to learn the graph and integrate the graph learning into the proposed l1-norm graph regularised optimisation problem.
- 13
- 12
- Video topic modelling with behavioural segmentation12
This paper proposes a novel region LDA model, which encodes the spatial awareness that is ignored by conventional topic models, and demonstrates the superiority of the proposed model on busy real world scenes.
- 11
- 11
- 11
- 11
- Learning Relevance Restricted Boltzmann Machine for Unstructured Group Activity and Event Understanding11
The experimental results demonstrate that the relevant features learned by the ReRBM provide better semantic and discriminative description for videos than a number of alternative supervised latent variable models, and achieves state of the art performance in terms of classification accuracy and retrieval precision.
- Learning spatio-temporal features for action recognition from the side of the video11
A novel spatio-temporal feature learning approach is introduced for action recognition that automatically detects and track the actor, and map the action track to a cuboid, which is then split into block sequences.
- Learning human pose in crowd11
A novel framework for locating people and inferring their body and head poses is proposed and significantly outperforms a state-of-the-art person detector and is able to yield extremely accurate head and body pose estimation in crowded public spaces.
- Post-Processing Temporal Action Detection10
This work introduces a novel model-agnostic post-processing method without model redesign and retraining, dubbed as Gaussian Approximated Post-processing (GAP), which enables lower temporal resolutions for more efficient inference, facilitating low-resource application.
- Directional region-based feature point matching algorithm based on SURF10
Experimental results show that the proposed algorithm can substantially improve the matching accuracy and the number of valid matched points, particularly in the presence of a large amount of noise and interference.
- DreamCreature: Crafting Photorealistic Virtual Creatures from Imagination9
This work proposes a new method called DreamCreature, which identifies and extracts the underlying sub-concepts in an unsupervised manner, and adapts to generate novel concepts with faithful structures and photorealistic appearance by seamlessly and flexibly composing learned sub- Concepts.
- 9
- 9
- 9
- 9
- Multi-Modal Few-Shot Temporal Action Detection8
A novel MUlti-modality PromPt mETa-learning (MUPPET) method that bridges pretrained vision and language models whilst maximally reusing already learned capacity to tackle few-shot object detection problem and achieves the state-of-the-art performance on MS-COCO dataset.
- Domain Attention Consistency for Multi-Source Domain Adaptation8
A novel approach: transferable attribute learning, which combines pseudo-labeling with a class compactness loss to minimize the distance between the target features and the classifier's weight vectors, and proposes a domain attention consistency network, dubbed DAC-Net.
- 8
- Surveillance video behaviour profiling and anomaly detection8
The results demonstrate that the framework is able to detect not only anomalies that are visually obvious, but also those that are ambiguous or supported only by very weak visual evidence, e.g. those that can be easily missed by a human observer.
- UIGR: Unified Interactive Garment Retrieval7
This paper proposes a Unified Interactive Garment Retrieval (UIGR) framework to unify TGR and VCR, and contributes a large-scale benchmark suited for both problems.
- 7
- 7
- A Global Occlusion-Aware Approach to Self-Supervised Monocular Visual Odometry7
A multi-scale non-local attention module, consisting of both intra-stage augmented attention and cascaded across-stage attention, is proposed for robust depth estimation given occlusion, alleviating the impacts of occlusions via global attention modeling.
- Behaviour in Context7
A range of established topics and emerging trends in visual analysis of behaviour from understanding facial expression, body gesture, action and intent, to the analysis of group activity, crowd and distributed behaviour, and gaining holistic awareness are introduced.
- Visual inference of human emotion and behaviour7
This work extends Canonical Correlation Analysis to multimodality correlation for bebaviour inference using both facial and body gestures and model temporal correlation among human movement patterns in a wider space using a mixture of Multi-Observation Hidden Markov Model for human behaviour profiling and behavioural anomaly detection.
- 6
- Random Forest with Adaptive Local Template for Pedestrian Detection6
This paper generates a few local templates with different sizes and different locations in positive exemplars and builds a Random Forest whose splitting functions are optimized by maximizing class purity of matching the local templates to the training samples, respectively to improve the classification accuracy.
- A Unified Bayesian Framework for Adaptive Visual Tracking6
A unified generative model for multi-sensory adaptive tracking which cleanly integrates tracking and the modeling of appearance change across multiple features in the same framework is derived.
- Single Stage Multi-Pose Virtual Try-On5
A novel single stage model for MPVTON with a parallel flow estimation module that predicts the flow fields for both person and garment images conditioned on the target pose is proposed, resulting in new SOTA performance on existing MPVton benchmarks.
- 5
- Discriminative boosted forest with convolutional neural network-based patch descriptor for object detection5
An object-detection method based on improved random forest and local image patches represented by CNN features that achieves state-of-the-art or competitive performance on three well-known challenging datasets.
- 5
- Unsupervised Selective Transfer Learning for Object Recognition5
A novel unsupervised transfer learning framework that utilises unlabelled auxiliary data to quantify and select the most relevant transferrable knowledge for recognising a target object class from the background given very limited training target samples is proposed.
- ConceptHash: Interpretable Fine-Grained Hashing via Concept Discovery4
Existing fine-grained hashing methods typically lack code interpretability as they compute hash code bits holistically using both global and local features, so this approach allows us to develop hash codes that exhibit similarity within families of species while remaining distinct from species in other families.
- Robust Target Training for Multi-Source Domain Adaptation4
A novel Bi-level Optimization based Robust Target Training (BORT) method for MSDA which turns any existing fully-trained one-step MSDA model into a labeling function to generate pseudo-labels for the target data and trains a target model using pseudo-labeled target data only.
- 4
- Deep Multi-View Learning with Stochastic Decorrelation Loss.4
A mini-batch based Stochastic Decorrelation Loss (SDL) is proposed which can be applied to any network layer to provide soft decorrelation of the layer's activations and is superior to other decorrelation losses in terms of efficacy and scalability.
- L1 Graph Based Sparse Model for Label De-noising4
This work proposes a novel robust graph-based approach for label de-noising, built upon a visual similarity graph in a form of L 1 graph regulariser, which is more robust against visual outliers than the conventional L 2 regulariser.
- Action recognition with cascaded feature selection and classification4
This work proposes a novel action classification approach based on cascaded feature selection and classification that significantly outperforms standard multi-class classifiers in addressing more challenging action recognition problems encountered in an unconstrained environment.
- Combined evaluation of audiology examination and self-reported symptoms in patients with hyperacusis3
The application of combined audiological examination and a self-reported symptoms survey in the evaluation of hyperacusis allows for an accurate and comprehensive assessment of hyperacusis.
- Smaller Is Bigger: Rethinking the Embedding Rate of Deep Hiding3
A novel Local Deep Hiding (LDH) scheme that significantly increases the embedding rate by hiding large secret images into small local regions of cover images and exhibits superior robustness to common image distortions.
- 3
- 3
- 3
- 3
- 2
- 2
- Generative Model Based Noise Robust Training for Unsupervised Domain Adaptation2
This paper proposes a Generative model-based Noise-Robust Training method (GeNRT), which eliminates domain shift while mitigating label noise, and gets theoretically linked to the Ben-David domain adaptation theorem for reducing the domain gap.
- Hyper-VolTran: Fast and Generalizable One-Shot Image to 3D Object Structure via HyperNetworks2
This work introduces a novel neural rendering technique that employs the signed distance function as the surface representation and incorporates generalizable priors through geometry-encoding volumes and HyperNetworks, and builds neural encoding volumes from generated multi-view inputs.
- Unsupervised Hashing with Similarity Distribution Calibration2
A novel Similarity Distribution Calibration (SDC) method is introduced, which aligns the hash code similarity distribution towards a calibration distribution with sufficient spread across the entire similarity capacity/range, to alleviate the similarity collapse problem.
- Learning to Augment via Implicit Differentiation for Domain Generalization2
This paper proposes a novel augmentation-based DG approach, dubbed AugLearn, which views a data augmentation module as hyper-parameters of a classification model and optimizes the module together with the model via meta-learning, and formulate an efficient joint training algorithm, for both the augmentation modules and the classification model, based on the implicit function theorem.
- 2
- Knowing What, Where and When to Look: Video Action modelling with Attention2
A novel What-Where-When (W3) video attention module is proposed, which is extremely efficient by factorising the high-dimensional video feature data into low-dimensional meaningful spaces (1D channel vector for ‘what’ and 2D spatial tensors for ‘where’), followed by a lightweight temporal attention reasoning.
- 2
- 2
- 2
- 2
- One-Shot Learning of Sketch Categories with Co-regularized Sparse Coding2
A transfer learning framework which enables one-shot learning of sketch categories and is based on a novel co-regularized sparse coding model which exploits common/shareable parts among human sketches of seen categories and transfer them to unseen categories.
- User-assisted visual search and tracking across distributed multi-camera networks2
A distributed multi-camera tracking system to visually aid human operators in associating people and objects effectively over multiple disjoint camera views in a large public space is described and an extensive evaluation of the greater effectiveness of the system as compared to existing approaches on industry-standard i-LIDS multi- camera data is provided.
- 1
- 1
- 1
- 1
- 1
- A Novel Vehicle Type Recognition Based on Hierarchical Matching1
The algorithm can recognize vehicle type in traffic scenes efficiently and realize the rapid vehicle type recognition through lookup tables by the hierarchical index of vehicle type template with k-means clustering and size normalization processing.
- 1
- Unsupervised Behaviour Profiling1
This chapter considers a clustering model that discovers the intrinsic grouping of behaviour patterns and detects anomalies on-line so that a decision on whether a behaviour pattern is normal is made as soon as sufficient visual evidence is collected without the completion of the observed pattern.
- Supervised Learning of Group Activity1
A contextual event based group activity representation, two different methods for activity based video segmentation, and a dynamic Bayesian network for supervised learning of an activity model with its model structure automatically learned from visual observations are described.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Robust Complex Behaviour Modeling at 90Hz–
A novel complex behaviour modeling framework that consists of a Binarized Cumulative Directional (BCD) feature as representation, novel spatial and temporal context modeling via an iterative correlation maximization, and a set of behaviour models, each being a simple Bernoulli distribution is introduced.
- Human Detection Based on a New Gradient CENTRIST Feature–
A real-time and accurate human detection method base on a new Gradient CENTRIST feature descriptor that can characterizes not only local human appearance and shape but also implicitly represent the global contour is proposed.
- People Counting Based on the Moving Feature of the Mass–
The people counting model is obtained by training a support vector machine (SVM) classifier with the moving feature and shape feature of mass, and the experimental results confirm that the approach improves the accuracy of people counting.
- –
- –
- Tensor Representations for Object Classification and Detection–
A novel method is proposed for robust detection and categori zation of people from a wide-area distance by appearance and action and a complete framework for pedestrian detection based on the SST combined with Hausdorff distance that is able to manage object’s descript ion of variable lengths is presented.
- –
- Hierarchical Behaviour Discovery–
This chapter describes a Markov clustering topic model designed for unsupervised modelling and on-line processing of multi-object spatio-temporal behaviours in crowded public scenes that draws on machine learning theories on probabilistic topic models and dynamic Bayesian networks to achieve a robust hierarchical modelling of behaviours and their dynamics.
- –
- Towards Modelling Behaviour–
This chapter introduces a statistical learning approach, in particular probabilistic graphical models, to underpinning the mechanism for behavioural interpretation, and overviews different learning strategies for building behaviour models, ranging from supervised learning, unsupervised learning, semisupervisedLearning, weakly supervised learning and active learning.
- –
- Man in the Loop–
This chapter describes an active learning model for seeking human feedback on model selected queries that is internal to the model rather than decided by the human observer, to ensure that not only the most relevant human feedback is sought, but also the model is not subject to human bias without data support.
- Quantum spin liquid phase in the spin-1 bilinear-biquadratic Heisenberg model on a honeycomb lattice–
- Understanding Facial Expression–
This chapter considers the problems of how to construct a suitable representation and design an effective classification model for both static image based and dynamic sequence based automatic facial expression recognition.
- Multi-camera Behaviour Correlation–
This chapter describes a framework for modelling a joined-up representation of a synchronised global space, within which local activities from different observational viewpoints can be analysed and interpreted holistically, and develops a suitable mechanism capable of discovering and quantifying unknown correlations in temporal ordering and temporal delays among different camera views.
- –
- Modelling Rare and Subtle Behaviours–
To eliminate the prohibitive manual labelling cost, both in time and inconsistency, required by traditional supervised methods, this chapter describes a weakly supervised framework, in which a user needs not, or cannot, explicitly locate the target behaviours of interest in the training video data.
- –
- Unsupervised learning of visual context using completed likelihood AIC–
This paper addresses the problem of choosing the most suitable probabilistic model selection criterion for learning visual context of a dynamic scene modelling with the aim of overcoming poor model selection by existing popular criteria when the data sample size varies from very small to large.
- On the distortion of shape recovery from motion–
A theoretical model is presented to understand the impact that errors in the motion estimates have on shape recovery, and it is shown that different shapes exhibit different degrees of robustness with respect to its recovery.
- –
- –
- Distortion of Shape from Motion–
A theoretical model is presented to understand the impact that errors in the motion estimates have on shape recovery and focuses on the recovery of second order shape under different generic motions, each of which presenting different degrees of error sensitivity.
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.