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
- Nuno M. VasconcelosSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works240 from public data
- 1,209
- 1,004
- 496
- 3D Human Pose Estimation from Monocular Images with Deep Convolutional Neural Network489
A deep convolutional neural network for 3D human pose estimation from monocular images is proposed and empirically show that the network has disentangled the dependencies among different body parts, and learned their correlations.
- 467
- 465
- 407
- Learning Dynamic Memory Networks for Object Tracking373
A dynamic memory network to adapt the template to the target's appearance variations during tracking, which can be easily enlarged as the memory requirements of a task increase, which is favorable for memorizing long-term object information.
- 249
- Heterogeneous Multi-task Learning for Human Pose Estimation with Deep Convolutional Neural Network248
A heterogeneous multi-task learning framework for human pose estimation from monocular images using a deep convolutional neural network and it is shown that including the detection tasks helps to regularize the network, directing it to converge to a good solution.
- Maximum-Margin Structured Learning with Deep Networks for 3D Human Pose Estimation237
This paper focuses on structured-output learning using deep neural networks for 3D human pose estimation from monocular images, and proposes an efficient recurrent neural network for performing inference with the learned image-embedding.
- 218
- 210
- Beyond Counting: Comparisons of Density Maps for Crowd Analysis Tasks—Counting, Detection, and Tracking209
Evaluating density maps generated by density estimation methods on a variety of crowd analysis tasks, including counting, detection, and tracking finds that the lower-resolution density maps sometimes have better counting performance, and proposes several metrics for measuring the quality of a density map.
- 195
- 182
- 181
- 173
- Understanding eye movements in face recognition using hidden Markov models171
It is found that correct and wrong recognitions were associated with distinctive eye movement patterns; the difference between the two patterns lies in the transitions rather than locations of the fixations alone.
- 151
- Scanpath modeling and classification with hidden Markov models149
A turnkey method for scanpath modeling and classification based on variational hidden Markov models (HMMs) and discriminant analysis (DA), which allow to integrate bottom-up, top-down, and oculomotor influences into a single model of gaze behavior.
- 147
- ROAM: Recurrently Optimizing Tracking Model145
To effectively adapt the model to appearance variations, this paper proposes to offline train a recurrent neural optimizer to update tracking model in a meta-learning setting, which improves the convergence speed of updating the tracking model while achieving better performance.
- A Comparative Survey of Deep Active Learning143
This work constructs a DAL toolkit, DeepAL+, by re-implementing 19 highly-cited DAL methods, and explores some factors that influence the efficacy of DAL, which provides better references for researchers to design their DAL experiments or carry out DAL-related applications.
- Automatic Musical Pattern Feature Extraction Using Convolutional Neural Network138
This work proposes a novel approach to extract musical pattern features in audio music using convolutional neural network (CNN), a model widely adopted in image information retrieval tasks.
- Modeling Noisy Annotations for Crowd Counting.121
Examination of the predictions of the trained model shows that it can correctly predict the locations of people in spite of the noisy training data, which demonstrates the robustness of the loss function to annotation noise.
- 121
- Analysis of Crowded Scenes using Holistic Properties120
A crowd counting system, based on regression of holistic (global) features, on the PETS 2009 dataset and experimental results on crowd event detection when using the dynamic texture model to represent holistic motion flow in the video are presented.
- 118
- Describing Like Humans: On Diversity in Image Captioning111
A new metric for measuring the diversity of image captions is proposed, derived from latent semantic analysis and kernelized to use CIDEr similarity, which shows that balancing the cross-entropy loss and C IDEr reward in reinforcement learning during training can effectively control the tradeoff between diversity and accuracy of the generated captions.
- 111
- Direct convex relaxations of sparse SVM108
Two direct, novel convex relaxations of a non-convex sparse SVM formulation that explicitly constrains the cardinality of the vector of feature weights are proposed that results in a quadratically-constrained quadratic program (QCQP) and a semidefinite programming (SDP) relaxation.
- 106
- Martial Arts, Dancing and Sports dataset: A challenging stereo and multi-view dataset for 3D human pose estimation103
The results of the evaluation suggest that discriminative approaches perform better than generative approaches when there are enough representative training samples, and that the generative methods are more robust to diversity of poses, but can fail to track when the motion is too quick for the effective search range of the particle filter.
- 97
- 97
- A Comparative Survey: Benchmarking for Pool-based Active Learning93
This paper surveys and compares various AL strategies used in both recently proposed and classic highly-cited methods, and proposes to benchmark pool-based AL methods with a variety of datasets and quantitative metric, and draws insights from the comparative empirical results.
- 92
- Incorporating Side Information by Adaptive Convolution88
An adaptive convolutional neural network (ACNN), where the convolution filter weights adapt to the current scene context via the side information, is proposed, which improves the performance compared to a plain CNN with a similar number of parameters and achieves similar or better than state-of-the-art performance on crowd counting task.
- Recurrent Filter Learning for Visual Tracking88
This paper directly feed the target's image patch to a recurrent neural network (RNN) to estimate an object-specific filter for tracking, and extends the matrix multiplications of the fully-connected layers of the RNN to a convolution operation on feature maps, which preserves thetarget's spatial structure and also is memory-efficient.
- 87
- 82
- Cross-View Cross-Scene Multi-View Crowd Counting81
A CVCS model that attentively selects and fuses multiple views together using camera layout geometry, and a noise view regularization method to train the model to handle non-correspondence errors is proposed.
- Hidden Markov model analysis reveals the advantage of analytic eye movement patterns in face recognition across cultures79
It is suggested that active retrieval of facial feature information through an analytic eye movement pattern may be optimal for face recognition regardless of culture.
- Directing user attention via visual flow on web designs78
This work builds two user attention models that characterize user attention patterns between a pair of page components that can effectively guide user attention through the web design according to the designer's high-level specification.
- 78
- Automatic stylistic manga layout78
This paper proposes an approach to automatically generate a stylistic manga layout from a set of input artworks with user-specified semantics, thus allowing less-experienced users to create high-quality manga layouts with minimal efforts.
- Generalized Stauffer–Grimson background subtraction for dynamic scenes78
An adaptive model for backgrounds containing significant stochastic motion (e.g. water) based on a generalization of the Stauffer–Grimson background model, where each mixture component is modeled as a dynamic texture is proposed.
- 76
- 68
- Visual Tracking via Dynamic Memory Networks65
A dynamic memory network to adapt the template to the target's appearance variations during tracking and design a “negative” memory unit that stores templates for distractors, which are used to cancel out wrong responses from the object template.
- 62
- A Lightweight and Detector-Free 3D Single Object Tracker on Point Clouds61
DMT is proposed, a Detector-free Motion-prediction-based 3D Tracking network that completely removes the usage of complicated 3D detectors and is lighter, faster, and more accurate than previous trackers.
- Objective Measures of is Usage Behavior Under Conditions of Experience and Pressure Using Eye Fixation Data61
The proposed statistical approach for analyzing eye-movement data is a critical methodological contribution to the emerging research that uses eye-tracking technology for investigation.
- 61
- 60
- A Family of Probabilistic Kernels Based on Information Divergence59
A family of probabilistic kernels is introduced, based on information divergence measures, that provide a unifying framework for the study of various important questions in kernel theory and practice and yield interesting results regarding the role of properties such as linearity, positive definiteness, and the triangle inequality.
- Eye movement analysis with hidden Markov models (EMHMM) with co-clustering57
This work proposed to combine EMHMM with the data mining technique co-clustering to discover participant groups with consistent eye-movement patterns across stimuli for tasks involving stimuli with different feature layouts and discovered explorative and focused eye- Movement patterns among Asian participants.
- Individuals with insomnia misrecognize angry faces as fearful faces while missing the eyes: an eye-tracking study57
The findings suggest that impaired information selection through visual attention control may be related to the compromised emotion perception in individuals with insomnia.
- Is having similar eye movement patterns during face learning and recognition beneficial for recognition performance? Evidence from hidden Markov modeling57
These findings suggested that perceptuomotor memory elicited by eye movement patterns during learning does not play an important role in recognition, and the retrieval of diagnostic information for recognition, such as the eyes for face recognition, is a better predictor for recognition performance.
- 57
- 56
- A Robust Panel Extraction Method for Manga56
This paper proposes a robust method for automatically extracting panels from digital manga pages by performing a recursive binary splitting to partition the panel block into a set of sub-blocks, where an optimal splitting line at each recursive level is determined adaptively.
- Dynamic Momentum Adaptation for Zero-Shot Cross-Domain Crowd Counting51
This work proposes a novel Crowd Counting framework built upon an external Momentum Template, termed C2MoT, which enables the encoding of domain specific information via an external template representation, and obtains state-of-the-art counting performance on the source domain.
- Scalable Video Object Segmentation with Simplified Framework50
A scalable Simplified VOS (SimVOS) framework to perform joint feature extraction and matching by leveraging a single transformer backbone for simultaneous feature extraction and matching between query and reference features is presented.
- 48
- Compare and Reweight: Distinctive Image Captioning Using Similar Images Sets48
A distinctiveness metric -- between-set CIDEr (CIDErBtw) is proposed to evaluate the distinctiveness of a caption with respect to those of similar images and shows that the human annotations of each image are not equivalent based on distinctiveness.
- Genre Classification and the Invariance of MFCC Features to Key and Tempo48
The invariance of MFCC to musical key and tempo is investigated, and it is shown that MFCCs in fact encode both timbral and key information, and that musical genres, which should be independent of key, are in fact influenced by the fundamental keys of the instruments involved.
- 44
- Look over here42
It is shown that the proposed approach to synthesize a composition of panel elements that can effectively guide the reader's attention to convey the story can measurably improve the readability, visual appeal, and communication of the story of the resulting pages, as compared to an existing method.
- 3D Crowd Counting via Geometric Attention-Guided Multi-view Fusion41
This work considers the variable height of the people in the 3D world and proposes to solve the multi-view crowd counting task through 3D feature fusion with 3D scene-level density maps, instead of the 2D density map on the ground-plane.
- Do portrait artists have enhanced face processing abilities? Evidence from hidden Markov modeling of eye movements41
It is found that portrait artists showed more eyes-focused eye movement patterns and outperformed novices in face matching, and participants' drawing rating was correlated with both eye movement pattern and performance, and artists' advantage in face processing is specific to tasks similar to their drawing experience such as face matching.
- 41
- Fine-Grained Crowd Counting40
This article proposes fine-grained crowd counting, which differentiates a crowd into categories based on the low-level behavior attributes of the individuals and then counts the number of people in each category.
- 40
- 40
- 40
- Color Orchestra: Ordering Color Palettes for Interpolation and Prediction39
A divide-and-conquer sorting algorithm is developed to rearrange the colors in the palettes in a coherent order, which allows meaningful interpolation between color palettes and leads to promising applications such as photo-style exploration, real-time color suggestion, and enriched photo recolorization.
- FlexyFont: Learning Transferring Rules for Flexible Typeface Synthesis39
FlexyFont is introduced, a flexible tool for synthesizing a complete typeface that has a consistent style with a given small set of glyphs and is able to predict the style of a novel typeface given only a few examples.
- RegGeoNet: Learning Regular Representations for Large-Scale 3D Point Clouds38
This paper presents a novel unsupervised neural architecture called RegGeoNet to parameterize an unstructured point set into a completely regular image structure dubbed as deep geometry image (DeepGI), such that spatial coordinates of unordered points are recorded in three-channel grid pixels.
- 37
- 35
- Leveraging Long-Term Predictions and Online Learning in Agent-Based Multiple Person Tracking34
Experimental results show that the tracking algorithm is suitable for predicting pedestrians' behaviors online without needing scene priors or hand-annotated goal information, and improves tracking in real-world crowded scenes under low frame rates.
- 32
- On Distinctive Image Captioning via Comparing and Reweighting32
A distinctiveness metric—between-set CIDEr (CIDErBtw) is proposed to evaluate the distinctiveness of a caption with respect to those of similar images and reveals that the human annotations of each image in the MSCOCO dataset are not equivalent based on distinctiveness.
- 32
- 31
- 31
- TWINS: A Fine-Tuning Framework for Improved Transferability of Adversarial Robustness and Generalization30
A novel statistics-based approach, Two-WIng NormliSation (TWINS)fine-tuning framework, which consists of two neural networks where one of them keeps the population means and variances of pre-training data in the batch normalization layers and is shown to be effective on a wide range of image classification datasets in terms of both generalization and robustness.
- 30
- Eye movement analysis with switching hidden Markov models29
EMSHMM reveals and provides quantitative measures of individual differences in cognitive behavior/style, making a significant impact on the use of eyetracking to study cognitive behavior across disciplines.
- 28
- Understanding visual attention to face emotions in social anxiety using hidden Markov models27
Eye Movement analysis with Hidden Markov Models revealed novel associations between eye-movement patterns and social anxiety symptoms that were not evident with standard analytical approaches, and can offer novel insights into psychopathology-related attention processes.
- Fusing Crowd Density Maps and Visual Object Trackers for People Tracking in Crowd Scenes27
A density fusion framework that fuses the Sparse Kernelized Correlation Filter with an estimated crowd density map using a convolutional neural network, yielding a refined response map that can significantly improves people tracking in crowd scenes and can be combined with other trackers to improve the tracking performance.
- 26
- Wide-Area Crowd Counting: Multi-view Fusion Networks for Counting in Large Scenes25
A deep neural network framework for multi-view crowd counting is proposed, which fuses information from multiple camera views to predict a scene-level density map on the ground-plane of the 3D world.
- Interpretation biases and visual attention in the processing of ambiguous information in chronic pain25
Theories propose that interpretation biases and attentional biases might account for the maintenance of chronic pain symptoms, but the interactions between these two forms of biases in the context of chronicPain are understudied.
- Edit Temporal-Consistent Videos with Image Diffusion Model24
This article proposes an elegant yet effective Temporal-Consistent Video Editing (TCVE) method to mitigate the temporal inconsistency challenge for robust text-guided video editing, and establishes a dedicated temporal Unet architecture to faithfully capture the temporal coherence of the input video sequences.
- 24
- 24
- The variational hierarchical EM algorithm for clustering hidden Markov models24
A variational hierarchical EM algorithm that clusters a given collection ofHMMs into groups of HMMs that are similar, in terms of the distributions they represent, and characterizes each group by a "cluster center", i.e., a novel HMM that is representative for the group.
- An Empirical Study on Distribution Shift Robustness From the Perspective of Pre-Training and Data Augmentation23
This paper studies the distribution shift problem from the perspective of pre-training and data augmentation, two important factors in the practice of deep learning that have not been systematically investigated by existing work.
- Efficient tree-structured SfM by RANSAC generalized Procrustes analysis22
A tree-structured structure-from-motion (SfM) method that recovers 3D scene structures and estimates camera poses from unordered image sets and is able to produce metric reconstructions efficiently and robustly is proposed.
- 22
- 22
- ODAM: Gradient-based instance-specific visual explanations for object detection21
The gradient-weighted Object Detector Activation Maps (ODAM), a visualized explanation technique for interpreting the predictions of object detectors, is proposed and higher-quality visual explanations are produced than the state-of-the-art both effectively and efficiently.
- Understanding the role of eye movement consistency in face recognition and autism through integrating deep neural networks and hidden Markov models21
The hypothesis that higher eye movement consistency across trials, instead of a greater eyes-focused pattern, predicts better performance in children since it reflects capacity in developing visual routines is tested.
- Group-based Distinctive Image Captioning with Memory Attention21
A Group-based Distinctive Captioning Model (GdisCap), which compares each image with other images in one similar group and highlights the uniqueness of each image, and a new evaluation metric, distinctive word rate (DisWordRate) to measure the distinctiveness of captions.
- 21
- 20
- Calibration-Free Multi-view Crowd Counting20
This paper proposes calibration-free multi-view crowd counting (CF-MVCC), which obtains the scene-level count directly from the density map predictions for each camera view without needing the camera calibrations in the test.
- 20
- Understanding the collinear masking effect in visual search through eye tracking20
It is suggested that participants' pre-saccadic attention shift prior to search may be an important factor influencing their search behavior, and collinear masking effect was associated with reduced eye-fixation consistency at the central fixation cross prior to the search display presentation.
- The effects of attentional and interpretation biases on later pain outcomes among younger and older adults: A prospective study19
The current study aims to fill research gaps by examining pain‐related cognitive biases in elderly persons through the lens of attentional biases and the role of interpretation biases.
- Gated Hierarchical Attention for Image Captioning19
This paper proposes a bottom-up gated hierarchical attention (GHA) mechanism for image captioning in which low- level concepts are merged into high-level concepts and simultaneously low-level attended features pass to the top to make predictions.
- 19
- 18
- Is that my hand? An egocentric dataset for hand disambiguation18
A dataset that simulates daily activities, with variable illumination and people from different cultures and ethnicity to address daily life conditions is proposed and three neural network architectures that jointly learn the hand and context information are proposed.
- The interrelation between interpretation biases, threat expectancies and pain‐related attentional processing17
The association between interpretation biases and pain‐related threat expectancies and their associations with eye movements to pain-related imagery is investigated.
- 16
- Grad-ECLIP: Gradient-Based Visual and Textual Explanations for CLIP14
A Gradient-based visual and textual Explanation method for CLIP (Grad-ECLIP), which interprets the matching result of CLIP for a specific input image-text pair and generates effective heat maps that reveal the impact of image regions or words on the CLIP results.
- 14
- Applying the Hidden Markov Model to Analyze Urban Mobility Patterns: An Interdisciplinary Approach14
An attempt to adopt a Hidden Markov Model (HMM) toolbox developed in Computer Science for the analysis of eye movement patterns in Psychology to answer urban mobility questions in Geography demonstrates an attempt to integrate spatial, temporal, and sequential information for understanding people’s behavior in urban mobility research.
- 13
- Threading Keyframe with Narratives: MLLMs as Strong Long Video Comprehenders13
Experimental results on multiple long-video benchmarks demonstrate that Nar-KFC significantly improves the performance of popular MLLMs, and serves as a temporal- and content-aware compression strategy that complements visual and textual modalities.
- DistinctAD: Distinctive Audio Description Generation in Contexts13
Comprehensive evaluations on MAD-Eval, CMD-AD, and TV-AD benchmarks demonstrate the superiority of DistinctAD, with the model consistently outperforming baselines, particularly in Recall@k/N, highlighting its effectiveness in producing high-quality, distinctive ADs.
- A Fixed-Point Approach to Unified Prompt-Based Counting13
This paper establishes a comprehensive prompt-based counting framework capable of generating density maps for concerned objects indicated by various prompt types, such as box, point, and text, by converting prompts from different modalities into prompt masks without requiring training.
- Mind reading: Discovering individual preferences from eye movements using switching hidden Markov models.13
Conference Theme: Integrating Psychological, Philosophical, Linguistic, Computational and Neural Perspectives.
- 13
- Visual attention to own‐ versus other‐race faces: Perspectives from learning mechanisms and task demands12
This work proposes to understand the effect and its contributing factors from the perspectives of learning mechanisms that involve joint learning of visual attention strategies and internal representations for faces, which can be modulated by quality of contact with other-race individuals including emotional and motivational factors.
- Understanding children's attention to traumatic dental injuries using eye‐tracking12
Preschool children noticed teeth, however, most of them did not have an attentional bias, implying that they did not interpret these TDI outcomes negatively, and only a few children avoided looking at images with TDIs indicating a potential negative impact.
- Understanding Children’s Attention to Dental Caries through Eye-Tracking12
Preschool children’s eye movement patterns and visual attention to images with and without dental caries via eye movement analysis using hidden Markov models (EMHMM) are explored and EMHMM is appropriate for assessing inter-individual differences in children's visual attention.
- 12
- Fully Nested Neural Network for Adaptive Compression and Quantization12
This work proposes a fully nested neural network (FN3) that runs only once to build a nested set of compressed/quantized models, which is optimal for different resource constraints and proposes an ordered dropout (ODO) operation that ranks the building blocks.
- Information Distribution Within Musical Segments12
Entropy analysis using two representative datasets from Wikifonia and the Essen folksong corpus shows that musical segments also have higher information content in segment beginnings than endings, and the highest information content was observed in the middle of the segments.
- 12
- Temporal Unlearnable Examples: Preventing Personal Video Data from Unauthorized Exploitation by Object Tracking11
A novel generative framework for generating Temporal Unlearnable Examples (TUEs) is proposed that achieves state-of-the-art performance in video dataprivacy protection, with strong transferability across VOT models, datasets, and temporal matching tasks.
- The Pitfalls and Promise of Conformal Inference Under Adversarial Attacks11
This study investigates the uncertainty of deep learning models by examining the performance of conformal prediction (CP) in the context of standard adversarial attacks within the adversarial defense community and proposed Uncertainty-Reducing AT (AT-UR) is validates the effectiveness.
- Robust Zero-Shot Crowd Counting and Localization With Adaptive Resolution SAM11
This paper proposes a simple yet effective crowd counting method by utilizing the Segment-Everything-Everywhere Model, an adaptation of the Segmentation Anything Model, to generate pseudo-labels for training crowd counting models, which achieves the best unsupervised performance in crowd counting, while also being comparable results to some supervised methods.
- 11
- 11
- Multi-View People Detection in Large Scenes via Supervised View-Wise Contribution Weighting10
A supervised view-wise contribution weighting approach is developed that better fuses multi-camera information under large scenes and is adopted to enhance the model's generalization ability and enable more practical evaluation and comparison.
- Eye movement analysis of children’s attention for midline diastema10
This study aimed to explore children and their educators’ perception of a midline diastema by applying eye-movement analysis using the hidden Markov models (EMHMM), finding that EMHMMs are valuable in analysing eye- Movement patterns in children and adults.
- BEV-Net: Assessing Social Distancing Compliance by Joint People Localization and Geometric Reasoning10
A multi-branch network, BEV-Net, is proposed to localize individuals in world coordinates and identify high-risk regions where social distancing is violated, and experiments on complex crowded scenes demonstrate the power of the approach and show superior performance over baselines derived from methods in the literature.
- 10
- 10
- 10
- 9
- Using statistics to search and annotate pictures: an evaluation of semantic image annotation and retrieval on large databases9
An extensive experimental evaluation of the supervised multi-class labeling (SML) model for semantic image annotation proposed by [16] establishes a unifying view of the performance of two classes of labeling and retrieval systems that were previously only evaluated under different experimental protocols.
- Single-Frame-Based Deep View Synchronization for Unsynchronized Multicamera Surveillance8
A synchronization model that works in conjunction with existing deep neural network (DNN)-based multiview models, thus avoiding the redesign of the whole model, is proposed and applied to different DNNs-based multicamera vision tasks under the unsynchronized setting, and achieve good performance compared to baselines.
- Mining probabilistic color palettes for summarizing color use in artwork collections8
A novel technique for summarizing the color use in large artwork collections is presented, based on a novel representation, probabilistic color palettes, which can intuitively summarize the contextual and stylistic use of colors in a collection of artworks.
- Analytic Eye Movement Patterns in Face Recognition are Associated with Better Performance and more Top-down Control of Visual Attention: an fMRI Study.8
Conference Theme: Integrating Psychological, Philosophical, Linguistic, Computational and Neural Perspectives.
- Video Individual Counting for Moving Drones7
The MovingDroneCrowd Dataset is introduced, featuring videos captured by fast-moving drones in crowded scenes under diverse illuminations, shooting heights and angles and proposes a Shared Density map-guided Network (SDNet) using a Depth-wise Cross-Frame Attention (DCFA) module to directly estimate shared density maps between consecutive frames.
- 7
- Boosting Adversarial Robustness From The Perspective of Effective Margin Regularization7
This paper investigates the scale-variant property of cross-entropy loss, which is the most commonly used loss function in classification tasks, and its impact on the effective margin and adversarial robustness of deep neural networks.
- 7
- 7
- 7
- Global and Local Priming Evoke Different Face Processing Strategies: Evidence From An Eye Movement Study7
The results suggest that the holistic pattern is linked to engagement of global attention/global information processing, whereas the analytic pattern represents the use of local attention/featural-based processing.
- 7
- Re-Attentional Controllable Video Diffusion Editing6
This paper proposes a Re-Attentional Controllable Video Diffusion Editing (ReAtCo) method and proposes an Invariant Region-guided Joint Sampling (IRJS) strategy to mitigate the intrinsic sampling errors and faithfully preserve the invariant region content with less border artifacts.
- ViSS-R1: Self-Supervised Reinforcement Video Reasoning6
This work introduces a novel self-supervised reinforcement learning GRPO algorithm (Pretext-GRPO) within the standard R1 pipeline, and proposes the ViSS-R1 framework, which streamlines and integrates pretext-task-based self-supervised learning directly into the MLLM's R1 post-training paradigm.
- 6
- 6
- Variational Nested Dropout6
A variational nested dropout (VND) operation that draws samples of multi-dimensional ordered masks at a low cost, providing useful gradients to the parameters of nested drop out is proposed, and a Bayesian nested neural network is designed that learns the order knowledge of the parameter distributions.
- 6
- 6
- Chinese White Dolphin Detection in the Wild6
This paper develops a practical system to detect Chinese White Dolphins in the wild automatically and incorporates the dolphin detector into a system prototype, which detects dolphins in video frames at 100 FPS per GPU with high accuracy.
- Hidden Markov model analysis reveals better eye movement strategies in face recognition6
Eye movement strategies that lead to better performance in face recognition with hidden Markov models (HMMs) can be revealed by analysis methods that do not take individual differences in both temporal and spatial dimensions of eye movements into account.
- Advancing Multiple Instance Learning with Continual Learning for Whole Slide Imaging5
This work provides a foundation for CL in large-scale, weakly annotated clinical datasets, paving the way for more adaptable and resilient diagnostic models.
- 5
- Pareto Optimization for Active Learning under Out-of-Distribution Data Scenarios5
A sampling scheme, Monte-Carlo Pareto Optimization for Active Learning (POAL), which selects optimal subsets of unlabeled samples with fixed batch size from the unlabeling data pool, and experimental results show its effectiveness on both classical Machine Learning (ML) and Deep Learning (DL) tasks.
- 5
- 5
- Neighbours Matter: Image Captioning with Similar Images5
An image captioning model based on KNN graphs composed of the input image and its similar images, where each node denotes an image or a caption is proposed, which is competitive with the models that employ scene graphs and image parsing trees.
- 5
- Eye Movement Pattern in Face Recognition is Associated with Cognitive Decline in5
Invited talk - Symposium D: Interdisciplinary approaches for detection in Alzheimer’s Disease.
- Eye movement pattern in face recognition is associated with cognitive decline in the elderly5
Invited talk - Symposium D: Interdisciplinary approaches for detection in Alzheimer’s Disease.
- 5
- That was fast! Speeding up NN search of high dimensional distributions.5
A novel and efficient algorithm for deciding whether to explore nodes during backtracking, based on a variational approximation, which reduces the number of computations per node, and overcomes the limitations of Bregman Ball Trees on high dimensional data.
- Understanding eye movements in face recognition with hidden Markov model5
It is shown that using HMMs, individuals’ eye movement strategies can be described with both fixation locations and transition probabilities, and it is found that correct and wrong recognitions were associated with distinctive eye movement strategies.
- Density-based Object Detection in Crowded Scenes4
Two new strategies are proposed, density-guided anchors (DGA) and density-guided NMS (DG-NMS), which uses object density maps to jointly compute optimal anchor assignments and reweighing, as well as an adaptive NMS.
- 4
- Group-Based Distinctive Image Captioning with Memory Difference Encoding and Attention4
A novel approach to enhance the distinctiveness of image captions, namely Group-based Differential Distinctive Captioning Method, which visually compares each image with other images in one similar group and highlights the uniqueness of each image.
- 4
- Hand Detection Using Zoomed Neural Networks4
This paper addresses hand detection and hand disambiguation which focuses on detecting left and right hands as different objects.
- 4
- 4
- Insomniacs misidentify angry faces as fearful faces because of missing the eyes: An eye-tracking study4
Eye movement examinations revealed that when recognizing angry faces, more insomniacs adopted an eye movement pattern focusing on the mouth while more controls adopted a pattern attending to both the eyes and the mouth suggesting that impaired information selection through visual attention control may account for the compromised emotion perception ininsomniac individuals.
- GeneQuery: A General QA-based Framework for Spatial Gene Expression Predictions from Histology Images3
GeneQuery is presented, which aims to solve this gene expression prediction task in a question-answering (QA) manner for better generality and flexibility and outperforms existing state-of-the-art methods on known and unseen genes.
- 3
- 3
- 3
- Parametric Manifold Learning of Gaussian Mixture Models3
This paper proposes Parametric Manifold Learning of GMMs (PMLGMM), which learns a parametric mapping from a low-dimensional latent space to a high-dimensional GMM manifold and demonstrates the effectiveness of PML-GMM through experiments on synthetic, eye-fixation, flow cytometry, and social check-in data.
- Optimal face recognition performance involves a balance between global and local information processing: Evidence from cultural difference.3
The results suggest that engagement of global/local attention leads to face-center/eye biased eye movements respectively, and optimal recognition performance involves both global and local processing/gaze transitions between the face center and eyes.
- Hidden Markov modeling of eye movements with image information leads to better discovery of regions of interest3
Conference Theme: Integrating Psychological, Philosophical, Linguistic, Computational and Neural Perspectives.
- Semi-Supervised Multi-View Crowd Counting by Ranking Multi-View Fusion Models2
Two semi-supervised multi-view crowd counting frameworks are proposed by ranking the multi-view fusion models of different numbers of input views, in terms of the model predictions or the model uncertainties, in terms of the model predictions or the model uncertainties.
- 2
- A Secure Image Watermarking Framework with Statistical Guarantees via Adversarial Attacks on Secret Key Networks2
This work proposes a novel framework to train a secret key network (SKN), which serves as a non-duplicable safeguard for securing the embedded watermark and demonstrates that the framework maintains robustness comparable to existing meth-ods and excels in security and imperceptibility.
- 2
- 2
- 2
- The EM Algorithm for Layered Dynamic Textures2
The EM algorithm is derived for the two models of layered dynamic textures that contains multiple state processes and hence can model video with multiple regions of dieren t motion.
- Learning Tracking Representations from Single Point Annotations1
A soft contrastive learning (SoCL) framework that incorporates target objectness prior into end-to-end contrastive learning and can achieve better performance than the fully supervised baseline trained with box annotations under the same annotation time cost.
- 1
- 1
- 1
- 1
- 1
- Understanding eye movements in face recognition with hidden Markov model - eScholarship1
A hidden Markov model (HMM)- based method to analyze eye movement data is proposed and it is shown that using HMMs, the authors can describe individuals’ eye movement strategies with both fixation locations and transition data using HMMs.
- –
- –
- Active View Selection for Scene-level Multi-view Crowd Counting and Localization with Limited Labeling Budget–
This paper proposes an independent view selection method (IVS) that considers view and scene geometries in the view selection strategy and conducts the view selection, labeling, and downstream tasks independently, and put forward an active view selection method (AVS) that jointly optimizes the view selection, labeling, and downstream tasks.
- –
- Regional crowd flow estimation from aerial view–
This paper presents a crowd flow estimation model whose goal is to estimate the flow into and out-of a certain region over any given time span and shows that it surpasses the performance of prior methods and can also be applied in a variety of crowd analysis applications for understanding social scenes.
- –
- –
- –
- –
- –
- –
- –
- –
- Hand disambiguation using attention neural networks in the egocentric perspective–
An Attention Network with various egocentric hand properties is used to address the computer vision task of detecting the hands and disambiguating them left from right, using the YOLO object detector and its Tiny version.
- –
- –
- EMHMM: Eye Movement Analysis with Hidden Markov Models and Its Applications in Cognitive Research.–
A novel eye movement data analysis method, Eye Movement analysis with Hidden Markov Models (EMHMM), which summarizes each individual’s eye movement pattern using a hidden Markov model (HMM; a type of machine learning model for time series data), including person-specific ROIs and transition probabilities among the ROIs.
- –
- –
- –
- –
- –
- Learning word embeddings via context grouping–
A grouping-based context predictive model is proposed by considering the interactions of context words, which generalizes the widely used CBOW model and Skip-Gram model and demonstrates better representations can be learned with suitable context groups.
- –
- –
- –
- Look Closely: Learning Exemplar Patches for Recognizing Textiles from Product Images–
This work develops a robust semi-supervised model that exploits both micro textures and macro deformable shapes to select representative patches from product images and shows that the patch selection method significantly improves the performance of conventional texture recognition methods that usually rely on dense sampling.
- Surveillance of Crowded Environments: Modeling the Crowd by Its Global Properties–
This chapter discusses the dynamic texture model for representing holistic motion flow, which treats the video as a sample from a linear dynamical system, and defines appropriate distances and kernels between dynamic textures so that crowd motion can be recognized with standard classification algorithms.
- –
- –
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.