Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks187 from public data
- 1,045
- UA-DETRAC: A new benchmark and protocol for multi-object detection and tracking799
This work performs a comprehensive quantitative study on the effects of object detection accuracy to the overall MOT performance, using the new large-scale University at Albany DETection and tRACking (UA-DETRAC) benchmark dataset.
- Multi-Scale Structure-Aware Network for Human Pose Estimation300
A robust multi-scale structure-aware neural network for human pose estimation that effectively improves state-of-the-art pose estimation methods that suffer from difficulties in scale varieties, occlusions, and complex multi-person scenarios.
- 169
- Variations of Dynamic Contrast-Enhanced Magnetic Resonance Imaging in Evaluation of Breast Cancer Therapy Response: A Multicenter Data Analysis Challenge150
The results suggest that the interalgorithm parameter variations are largely systematic, which are not likely to significantly affect the utility of DCE-MRI for assessment of therapy response.
- 145
- Spatio-Temporal Phrases for Activity Recognition135
This paper proposes an approach that efficiently identifies both local and long-range motion interactions and can capture the combination of the hand movement of one person and the foot response of another person, the local features of which are both spatially and temporally far away from each other.
- Parallel Residual Bi-Fusion Feature Pyramid Network for Accurate Single-Shot Object Detection125
A new parallel FP structure with bi-directional (top-down and bottom-up) fusion and associated improvements to retain high-quality features for accurate localization and achieves state-of-the-art performance on the UAVDT17 and MS COCO datasets.
- 119
- SMILEtrack: SiMIlarity LEarning for Occlusion-Aware Multiple Object Tracking116
This paper proposes an SLM that calculates the appearance similarity between two objects, overcoming the limitations of feature descriptors in Separate Detection and Embedding models, and develops a Similarity Matching Cascade (SMC) module with a novel GATE function for robust object matching across consecutive video frames, further enhancing MOT performance.
- Eyes Tell All: Irregular Pupil Shapes Reveal GAN-Generated Faces105
It is shown that GAN-generated faces can be exposed via irregular pupil shapes, caused by the lack of physiological constraints in the GAN models, and an automatic method to segment the pupils from the eyes and analyze their shapes is designed to distinguish GANs from real ones.
- The 4th AI City Challenge102
The 4th annual edition of the AI City Challenge has attracted 315 participating teams, who leverage city-scale real traffic data and high-quality synthetic data to compete in four challenge tracks, and results show promise that AI technology can enable smarter and safer transportation systems.
- The 5th AI City Challenge101
The fifth AI City Challenge attracted 305 participating teams across 38 countries, who leveraged city-scale real traffic data and high-quality synthetic data to compete in five challenge tracks, showing the promise of AI in Smarter Transportation.
- GAN-Generated Faces Detection: A Survey and New Perspectives95
This work aims to provide a comprehensive review of recent progress in GAN-face detection, focusing on methods that can detect face images that are generated or synthesized from GAN models.
- 90
- 89
- The 2019 AI City Challenge.85
Participation in this challenge has grown five-fold this year as tasks have become more relevant to traffic optimization and challenging to the computer vision community.
- 85
- 82
- FishEye8K: A Benchmark and Dataset for Fisheye Camera Object Detection70
An open FishEye8K benchmark dataset for road object detection tasks, which comprises 157K bounding boxes across five classes (Pedestrian, Bike, Car, Bus, and Truck), and benchmark results of State-of-The-Art (SoTA) models, including variations of YOLOv5, YOLOR, Y OLO7, and YolOv8 are presented.
- 70
- 69
- 69
- Multi-Camera Multi-Target Tracking with Space-Time-View Hyper-graph67
A new multi-camera multi-target tracking method based on a space-time-view hyper-graph that encodes higher-order constraints on 3D geometry, appearance, motion continuity, and trajectory smoothness among 2D tracklets within and across different camera views is proposed.
- The 8th AI City Challenge63
The eighth AI City Challenge highlighted the convergence of computer vision and artificial intelligence in areas like retail, warehouse settings, and Intelligent Traffic Systems, presenting significant research opportunities, with participants setting new benchmarks.
- DETRAC: A New Benchmark and Protocol for Multi-Object Tracking63
This work performs a comprehensive quantitative study on the effect of object detection accuracy to the overall MOT performance based on a new large-scale DETection and tRACking (DETRAC) benchmark dataset.
- 63
- 59
- Robust Attentive Deep Neural Network for Detecting GAN-Generated Faces57
This work proposes a robust, attentive, end-to-end framework that spots GAN-generated faces by analyzing eye inconsistencies, and automatically learns to identify inconsistent eye components by localizing and comparing artifacts between eyes.
- 52
- 51
- Multi-Scale Supervised Network for Human Pose Estimation45
A robust pose estimation method based on the recent deep conv-deconv modules with two improvements: (1) multi -scale supervision of body keypoints, and (2) a global regression to improve structural consistency of keypoints.
- 44
- UU-Mamba: Uncertainty-aware U-Mamba for Cardiac Image Segmentation40
A new UU-Mamba model is proposed, integrating the U-Mamba model with the Sharpness-Aware Minimization (SAM) optimizer and an uncertainty-aware loss function, which combines region-based, distribution-based, and pixel-based loss designs to improve segmentation accuracy and robustness.
- PreAugNet: improve data augmentation for industrial defect classification with small-scale training data40
A novel off-line pre-augmentation network (PreAugNet) which acts as a class boundary classifier that can effectively screen the quality of the augmented samples and improve image augmentation is introduced.
- 39
- Measuring 3D shape similarity by graph-based matching of the medial scaffolds39
This work proposes to measure 3D shape similarity by matching a medial axis (MA) based representation-the medial scaffold(MS), which approximates the theoretical optimal deformation path between two shapes by modeling shape deformations as discrete topological changes of the MS hypergraphs.
- 31
- A skull stripping method using deformable surface and tissue classification31
The algorithm takes on average 2 minutes to run on a typical PC and produces a brain mask and membership functions for gray matter, white matter, and cerebrospinal fluid and is robust to inhomogeneous intensity often observed in magnetic resonance images.
- A New Benchmark and Model for Challenging Image Manipulation Detection30
A new Challenging Image Manipulation Detection (CIMD) benchmark dataset is introduced and a new two-branch network model based on HRNet is proposed that can better detect both the image-editing and compression artifacts in those challenging conditions.
- A machine learning inversion scheme for determining interaction from scattering30
This paper shows that the effective potential can be probabilistically inferred from the scattering spectra without any restriction imposed by model assumptions, and can effectively enable quantification of interaction in highly correlated systems using scattering and diffraction experiments.
- The 9th AI City Challenge29
The ninth AI City Challenge continues to advance real-world applications of computer vision and AI in transportation, industrial automation, and public safety, and saw a 17% increase in participation, with 245 teams from 15 countries registered on the evaluation server.
- Quantitative pharmacokinetic analysis of prostate cancer DCE-MRI at 3T: comparison of two arterial input functions on cancer detection with digitized whole mount histopathological validation29
DCE-MRI using both AIF methods is excellent in discriminating PCa from NNPT, and if quantitative DCE-MRI is to be used as a biomarker in PCa, the same AIF method should be used consistently throughout the study.
- MixNet: Toward Accurate Detection of Challenging Scene Text in the Wild26
MixNet, a hybrid architecture that combines the strengths of CNNs and Transformers, capable of accurately detecting small text from challenging natural scenes, is presented, regardless of the orientations, styles, and lighting conditions.
- AI City Challenge 2019 -- City-Scale Video Analytics for Smart Transportation25
A new multi-camera fusion method is developed by extending the state-of-the-art single-camera tracking-by-detection with site calibrations and a Pyramid Granularity Attentive Model for ReID is pro-posed by improving the recent Region-Aware deep Model with a pyramid design and training strategy improvements.
- 24
- 23
- 23
- PseudoProp: Robust Pseudo-Label Generation for Semi-Supervised Object Detection in Autonomous Driving Systems22
This paper proposes a new approach, Pseudo-Prop, to generate robust pseudo-labels by leveraging motion continuity in video frames by using a novel bidirectional pseudo-label propagation approach to compensate for misdetection.
- 22
- MagGAN: High-Resolution Face Attribute Editing with Mask-Guided Generative Adversarial Network20
This work presents Mask-guided Generative Adversarial Network (MagGAN) for high-resolution face attribute editing, in which semantic facial masks from a pre-trained face parser are used to guide the fine-grained image editing process.
- Vocbench: A Neural Vocoder Benchmark for Speech Synthesis19
VocBench is presented, a framework that benchmark the performance of state-of-the-art neural vocoders and demonstrates that the framework can show competitive efficacy and quality of the synthesized samples for each vocoder.
- Learnable Discrete Wavelet Pooling (LDW-Pooling) for Convolutional Networks19
The Learning Discrete Wavelet Pooling (LDW-Pooling) is introduced that can be applied universally to replace standard pooling operations to better extract features with improved accuracy and efficiency.
- 19
- 18
- Deep learning-based super-resolution for small-angle neutron scattering data: attempt to accelerate experimental workflow18
An alternative route to circumvent the limitation of neutron flux using the recent deep learning super-resolution technique is proposed, which will allow users to make critical decisions at a much earlier stage of data collection, which can accelerate the overall experimental workflow.
- 18
- Open-Eye: An Open Platform to Study Human Performance on Identifying AI-Synthesized Faces17
An online platform called Open-eye is developed to study the human performance of Al-synthesized faces detection and the design and workflow of the Open- Eye is described.
- Efficient algorithms for graph regularized PLSA for probabilistic topic modeling17
This work describes a method for probabilistic topic analysis in image and text based on a new representation of graph-regularized PLSA (GPLSA), and proposes efficient multiplicative iterative algorithms for GPLSA with three popular regularizers, namely l1, l2 and symmetric KL divergences.
- FAST SEARCH ALGORITHMS FOR INDUSTRIAL INSPECTION17
An efficient general purpose search algorithm for alignment and an applied procedure for IC print mark quality inspection that is robust with respect to linear change of image intensity and thus can be applied to general industrial visual inspection.
- 16
- 16
- 15
- 14
- GPU Acceleration for FHEW/TFHE Bootstrapping14
This research focuses on designing and implementing a Graphic Processing Unit (GPU)-based accelerator for the third generation FHEW/TFHE bootstrapping scheme, which features smaller parameters and bootstrapping keys particularly suitable for GPU architectures compared to the other generations.
- 14
- 14
- 14
- Training-Free Image Manipulation Localization Using Diffusion Models13
A training-free IML approach using diffusion models that adaptively selects an appropriate number of diffusion timesteps for each input image in the forward process and performs both conditional and unconditional reconstructions in the backward process without relying on external conditions.
- 13
- 12
- 12
- Scattering-based structural inversion of soft materials via Kolmogorov–Arnold networks11
This work presents a machine learning framework based on the Kolmogorov-Arnold Network for directly extracting real-space structural information from scattering spectra in reciprocal space, and demonstrates its ability to accurately and efficiently resolve structural collectivity and complexity.
- 11
- 11
- Semi-supervised learning for topographic map analysis over time: a study of bridge segmentation11
A semi-supervised learning method that can transfer the annotation of maps across years and allow map comparison and anthropogenic studies across time is proposed and is capable of detecting target features (bridges) on historical maps without annotations.
- 11
- Multi-Teacher Single-Student Visual Transformer with Multi-Level Attention for Face Spoofing Detection11
A new Multi-Teacher Single-Student (MTSS) visual Transformer with a multi-level attention design is proposed with a novel Multi-Level Attention Module with a DropBlock (MAMD) designed to strengthen discriminative features while dropping irrelevant spatial features to avoid overfitting.
- Occupancy Sensing and Activity Recognition with Cameras and Wireless Sensors11
It is found that packet loss due to wireless interference has a significant effect on time series classification and the convolutional neural networks significantly outperforms the conventional support vector machine method.
- 11
- 10
- Image classification for historical documents: a study on Chinese local gazetteers9
The results demonstrate the effectiveness of deep learning models in accurately recognizing and categorizing historical local gazetteer illustrations and developed a user-friendly web service to enable researchers easy access to the developed models.
- 9
- 9
- 9
- Spotting the Fakes: A Deep Dive into GAN-Generated Face Detection8
This study offers a comprehensive review of recent advancements in GAN-generated face detection, focusing on techniques that detect facial images generated by GAN models.
- Inferring effective electrostatic interaction of charge-stabilized colloids from scattering using deep learning8
The results clearly indicate that a deep learning solution offers a reliable and quantitative method for studying molecular interactions and demonstrates the potential of deep learning techniques in tackling complex problems in soft-matter structures and beyond.
- 8
- 8
- Who Brings the Frisbee: Probing Hidden Hallucination Factors in Large Vision-Language Model via Causality Analysis7
This study proposes a novel causal approach: a hallucination probing system to identify hidden factors that induce hallucination, and analyzes the causality between images, text prompts, and network saliency to block these factors.
- 7
- 7
- TransRPN: Towards the Transferable Adversarial Perturbations using Region Proposal Networks and Beyond7
As RPNs have no constraint on size of input image, the TransRPN method can generate the adversarial images directly into object detectors with arbitrary input size, which improves the feasibility of the method in practical applications.
- 7
- 7
- Fast Online Upper Body Pose Estimation from Video7
This work describes a fast online video upper body pose estimation method (CDBN-MODEC) that is based on a conditional dynamic Bayesian network model, which predicts upper body poses in a frame without using information from future frames.
- 6
- A Comprehensive Review of Machine Learning Advances on Data Change: A Cross-Field Perspective6
A three-phase problem categorization scheme is proposed to link the key ideas in the two technical fields, domain shift and concept drift, to provide a novel scope for researchers to explore contemporary technical strategies, learn industrial applications, and identify future directions for addressing data change challenges.
- 6
- 6
- 6
- 3D Single-Person Concurrent Activity Detection Using Stacked Relation Network6
Stacked Relation Networks (SRN) is proposed, with a specialized relation network for each decomposition, so as to enhance the expressiveness of instance-wise representations via the inter-instance relationship modeling.
- Modeling Latent Comorbidity for Health Risk Prediction Using Graph Convolutional Network.6
The hypothesis is that patient health trend, disease prognosis, treatment, and their latent correlations can all be modeled by recovering the missing links in this bipartite graph (the link prediction problem) to lead to robust prediction across individual patients and health records.
- 6
- Reliable fusion of knee bone laser scans to establish ground truth for cartilage thickness measurement6
This work reduces the alignment error of point-to-point ICP by 30% and reduced coefficient of variation of cartilage thickness measurements from 5% down to 1.4%, significantly improving the method's repeatability.
- 6
- Improving Limited Supervised Foot Ulcer Segmentation Using Cross-Domain Augmentation Strategies5
A cross-domain augmentation method named TransMix is proposed that combines Augmented Global Pre-training (AGP) and Localized CutMix Fine-tuning (LCF) to enrich wound segmentation data for model learning to effectively improve the foot ulcer segmentation model training.
- MOTE-NAS: Multi-Objective Training-based Estimate for Efficient Neural Architecture Search5
Multi-Objective Training-based Estimate (MOTE) is developed, designed to model the actual performance of DNNs from macro to micro perspective by drawing the loss landscape and convergence speed simultaneously and delivering a model more accurate than KNAS.
- 5
- 5
- 5
- 5
- 5
- Automated determination of arterial input function for DCE-MRI of the prostate5
This paper proposes a fully automated method for determining the AIF from prostate DCE-MRI based on modeling pixel uptake curves as gamma variate functions (GVF), which obtained promising detection accuracy, and a very good match with expert traced manual AIF.
- Subvoxel segmentation and representation of brain cortex using fuzzy clustering and gradient vector diffusion5
An automatic and fast algorithm to segment the brain cortex and to represent it as a geometric surface on which analysis can be carried out is proposed, which works on T1 weighted MR brain images with extracranial tissue removed.
- 4
- 4
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention4
The technical contributions of this work include a hybrid design named Intra-Block Fusion (IBF) to strengthen the extracted features within each convolution block; and a novel Cross-Scale Attention (CSA) module to mitigate the scaling inconsistencies arising from the limited training samples, especially for cross-domain tasks.
- 4
- 4
- 4
- Deep intelligent network for device-free people tracking4
This work proposes to use deep neural networks together with visual tracking systems to effectively generate training data so as to learn a general model for RF sensor network measurements without the need for manual RF model parameter tuning.
- 4
- Co-Regularized PLSA for Multi-Modal Learning4
This work derives efficient iterative learning algorithms for coPLSA with symmetric KL, L2 and L1 divergences as co-regularizers, in each case the essential optimization problem affords simple numerical solutions that entail only matrix arithmetic operations and numerical solution of 1D nonlinear equations.
- Integrating manual preprocessing with automated feature extraction for improved rodent seizure classification3
The feasibility of video-based seizure stage detection and classification for rodent models of temporal lobe seizures is demonstrated using a semi-automatic pipeline that requires manual preprocessing of data, which underscores the ability to undertake quantitative analysis of rodent behavior.
- Optimizing Encrypted Neural Networks: Model Design, Quantization and Fine-Tuning Using FHEW/TFHE3
An FHE-aware Quantization and Fine-tuning framework that optimizes model parameters to align with FHE-Neuron’s constraints, ensuring high accuracy in encrypted inference in Deep Neural Network inference.
- Unsupervised domain adaptation for cross-style, cross-year land use understanding from historical maps3
A new Map Style Segmentation (MapStyleSeg) method that employs unsupervised domain adaptation (UDA) from deep learning (DL) to enhance cross-style, cross-year automatic map segmentation and conversion to address the challenges of digitizing historical maps with varying styles.
- 3
- 3
- Instance Contour Adjustment via Structure-Driven CNN3
The meaningfulness of image textures with clear and user-desired contours are guaranteed by the structure-driven CNN, which does not require any semantic label as input, which ensures its well generalization capa-bility.
- 3
- 2
- 2
- 2
- 2
- Driver License Field Detection Using Real-Time Deep Networks2
An automatic system for real-time visual detection and recognition of multiple driver’s license fields using an effective deep YOLOv3 detection network and this approach does not rely on text detection or OCR and outperforms them when tested with large viewing angles.
- Cross-cultural Training Analysis Via Social Science and Computer Vision Methods2
A joint analysis of military cross-cultural training data by Computer Vision specialists from GE Global Research as well as analyses from Georgetown University's Social Interaction Research Group are reported on.
- 2
- Fast Search Algorithms for IC Printed Mark Quality Inspection12
An effective and general purpose search algorithm for alignment is presented and applied to IC printed mark quality inspection and achieves high accuracy, reliability, and repeatability with high speed for industrial requirement and works well on field test of various IC products.
- Partial Ring Scan: Revisiting Scan Order in Vision State Space Models1
Partial RIng Scan Mamba (PRISMamba), a rotation-robust traversal that partitions an image into concentric rings, performs order-agnostic aggregation within each ring, and propagates context across rings through a set of short radial SSMs is presented.
- PatchEAD: Unifying Industrial Visual Prompting Frameworks for Patch-Exclusive Anomaly Detection1
This work proposes a unified patch-focused framework, Patch-Exclusive Anomaly Detection (PatchEAD), enabling training-free anomaly detection that is compatible with diverse foundation models, and constructs visual prompting techniques, including an alignment module and foreground masking.
- 1
- 1
- 1
- Dual-Process Scaffold Reasoning for Enhancing LLM Code Debugging1
A novel psychologically backed Scaffold Reasoning framework for code debugging, which encompasses the Scaffold Stream, Analytic Stream, and Integration Stream, which achieves an 88.91% pass rate and an average inference time of 5.36 seconds per-problem on DebugBench, outperforming other reasoning approaches across various LLMs in both reasoning accuracy and efficiency.
- 1
- Scale-Aware Crowd Count Network with Annotation Error Correction1
A Scale-Aware Crowd Counting Network (SACC-Net) is introduced that introduces a ``scale-aware'' architecture with error-correcting capabilities of noisy annotations, leading to improved convergence efficiency with comparable accuracy.
- 1
- 1
- 1
- 1
- –
- Continual Learning with Support Boundary Experience Blending–
Support Boundary Data (SBD), generated via differential-privacy-inspired noise into latent features to create boundary-adjacent representations that implicitly regularize decision boundaries, is introduced, leading to more stable and robust continual learning.
- –
- Uncertainty-Aware Proxy Attribute Reasoning for Reliable Media Bias Detection–
An uncertainty-aware proxy reasoning (UAPR) method that integrates proxy attribute prediction with uncertainty estimation to assess and stratify reasoning quality and shows a 6.25% relative F1 improvement over competitive baselines, demonstrating that uncertainty-aware reasoning boosts performance and provides insight into decision making.
- –
- –
- –
- –
- –
- DuoAD: Leveraging [CLS] Dual Characteristics for Training-Free Few-Shot Anomaly Detection–
This work identifies the dual characteristics of the ViT [CLS] token: its embedding provides anomaly-invariant global semantic representation, while its attention maps implicitly highlight spatially abnormal regions.
- –
- –
- –
- –
- –
- LSSGen: Leveraging Latent Space Scaling in Flow and Diffusion for Efficient Text to Image Generation–
- –
- –
- –
- Special issue on smart city technologies and services based on AI for digital twin applications–
HSP-VTON, a framework that combines a refined two-stage semantic segmentation approach to improve the mask precision with a Latent Consistency Model that accelerates diffusion-based image generation, addresses the dual challenges of achieving high-quality garment alignment and reducing the computational cost.
- –
- –
- –
- –
- –
- Addressing Long-Tail Noisy Label Learning Problems: a Two-Stage Solution with Label Refurbishment Considering Label Rarity–
This work introduces an effective two-stage approach by combining soft-label refurbishing with multi-expert ensemble learning, providing a principled solution to the long-tail noisy label problem.
- –
- –
- –
- –
- Scale-Aware Crowd Counting Using a Joint Likelihood Density Map and Synthetic Fusion Pyramid Network–
This work develops a Synthetic Fusion Pyramid Network (SPF-Net) with a scale-aware loss function design for accurate crowd counting and can accurately predict the locations of people in the crowd, despite training on noisy training annotations.
- Class-Specific Channel Attention for Few-Shot Learning–
Class-Specific Channel Attention (CSCA) module is proposed, which learns to highlight the discriminative channels in each class by assigning each class one CSCA weight vector, and aims to learn local and class-specific features with very effective computation.
- –
- –
- –
- A General Approach to Model Biomedical Data from 3D Unorganised Point Clouds with Medial Scaffolds–
The radius field of the MS provides an intuitive definition for morphological measurements, while the graph structure made explicit by the MS is useful for shape registration and matching applications, which are ready to use in a number of practical applications.
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 2026-10-15. 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.