A5072751690- Video Surveillance and Tracking Methods
- Infrared Target Detection Methodologies
- Autonomous Vehicle Technology and Safety
- Visual Attention and Saliency Detection
- Advanced Steganography and Watermarking Techniques
- Anomaly Detection Techniques and Applications
- Human Pose and Action Recognition
- Digital Media Forensic Detection
- Image Enhancement Techniques
- Semantic Web and Ontologies
- Fire Detection and Safety Systems
- Internet Traffic Analysis and Secure E-voting
- Service-Oriented Architecture and Web Services
- Advanced Vision and Imaging
- Chaos-based Image/Signal Encryption
- Impact of Light on Environment and Health
- Human-Animal Interaction Studies
- Business Process Modeling and Analysis
- Domain Adaptation and Few-Shot Learning
- Advanced Image and Video Retrieval Techniques
The University of Sydney
2022-2024
Beihang University
2018
In this paper, we propose a unified Multi-Object Tracking (MOT) framework learning to make full use of long term and short cues for handling complex cases in MOT scenes. Besides, better association, switcher-aware classification (SAC), which takes the potential identity-switch causer (switcher) into consideration. Specifically, proposed includes Single Object (SOT) sub-net capture cues, re-identification (ReID) extract classifier matching decisions using extracted features from main target...
Unsupervised learning has been popular in various computer vision tasks, including visual object tracking. However, prior unsupervised tracking approaches rely heavily on spatial supervision from templatesearch pairs and are still unable to track objects with strong variation over a long time span. As unlimited self-supervision signals can be obtained by video along cycle time, we investigate evolving Siamese tracker videos forward-backward. We present novel framework, which learn temporal...
Process-supervised reward models serve as a fine-grained function that provides detailed step-wise feedback to model responses, facilitating effective selection of reasoning trajectories for complex tasks. Despite its advantages, evaluation on PRMs remains less explored, especially in the multimodal domain. To address this gap, paper first benchmarks current vision large language (VLLMs) two types models: output (ORMs) and process (PRMs) multiple vision-language benchmarks, which reveal...
Large scale surveillance video analysis is one of the most important components in future artificial intelligent city. It a very challenging but practical system, consists multiple functionalities such as object detection, tracking, identification and behavior analysis. In this paper, we try to address three tasks hosted NVIDIA AI City Challenge contest. First, system that transforming image coordinate world has been proposed, which useful estimate vehicle speed on road. Second, anomalies...
There are many decisions which usually made heuristically both in single object tracking (SOT) and multiple (MOT). Existing methods focus on tackling decision-making problems special tasks without a unified framework. In this paper, we propose decision controller (DC) is generally applicable to SOT MOT tasks. The learns an optimal policy with deep reinforcement learning algorithm that maximizes long term performance supervision. To prove the generalization ability of DC, apply it challenging...
Joint detection and tracking, which solves two fundamental vision challenges in a unified manner, is challenging topic computer vision. In this area, the proper use of spatial-temporal information videos can help reduce local defects improve quality feature representations. Although modeling low-level (usually pixel-wise) has been studied, instance-level correlations (i.e., relations between semantic regions instances have occurred) not fully exploited. comparison, correlation more flexible...
AIGC (AI-Generated Content) has achieved tremendous success in many applications such as text-to-image tasks, where the model can generate high-quality images with diverse prompts, namely, different descriptions natural languages. More surprisingly, emerging personalization techniques even succeed describing unseen concepts only a few personal references, and there have been some commercial platforms for sharing valuable personalized concept. However, an advanced technique also introduces...
Diffusion models have achieved remarkable success in generating high-quality images. Recently, the open-source represented by Stable (SD) are thriving and accessible for customization, giving rise to a vibrant community of creators enthusiasts. However, widespread availability customized SD has led copyright concerns, like unauthorized model distribution unconsented commercial use. To address it, recent works aim let output watermarked content post-hoc forensics. Unfortunately, none them can...
Multi-Object Tracking (MOT) is a popular topic in computer vision. However, identity issue, i.e., an object wrongly associated with another of different identity, still remains to be challenging problem. To address it, switchers, confusing targets thatmay cause issues, should focused. Based on this motivation,this paper proposes novel switcher-aware framework for multi-object tracking, which consists Spatial Conflict Graph model (SCG) and Switcher-Aware Association (SAA). The SCG eliminates...