.png){kind=icon}
返回
FETrack: One-Stream Framework-Based Feature Enhancement for Object Tracking
.png){kind=icon}
PRE
AI
DOI：10.1016/j.dsp.2026.105935.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
• Novel Global Enhancement (GE) and Cross-Depth Template Fusion (CDTF) modules: the GE suppresses background interference, while the CDTF ensures robust contextual propagation, jointly enhancing feature representation. • An unsupervised hard sample learning strategy leveraging contrastive learning significantly augments feature discriminability by treating each candidate token as an independent instance. • A distillation-based fine-tuning approach optimizes parameters of multiple integrated modules in the backbone, ensuring their synergistic collaboration with the original architecture. • Comprehensive experiments on six public benchmarks demonstrate FETrack’s effectiveness, confirm its state-of-the-art performance, and validate the transferability of its proposed approaches to enhance other one-stream trackers.
期刊
D
IF:
3
论文数:
785
被引数:
0
机构
引用论文
Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks更快的r-cnn: 基于区域建议网络的实时目标检测
OAAI
OAAI分享收藏Fine-tuning feature interaction for unsupervised domain adaptive low-light object detection微调特征交互以实现无监督域自适应低光目标检测.png)
Neurocomputing
NeurocomputingIF6.5
PREAI
分享收藏Distilled large language model-driven dynamic sparse expert activation mechanism蒸馏大语言模型驱动的动态稀疏专家激活机制
PREAI
分享收藏Wu, Y.; Li, Y.; Liu, M.; Wang, X.; Yang, X.; Ye, H.; Zeng, D.; Zhao, Q.; Li, S. Learning an Adaptive and View-Invariant Vision Transformer for Real-Time UAV Tracking. arXiv 2024, arXiv:2412.20002. [Google Scholar] [CrossRef]Wu, Y.; Li, Y.; Liu, M.; Wang, X.; Yang, X.; Ye, H.; Zeng, D.; Zhao, Q.; Li, S. 学习一种自适应且视图不变的视觉Transformer，用于实时UAV跟踪。arXiv 2024, arXiv:2412.20002. [Google Scholar] [CrossRef]
IF0
PREAI
分享收藏Efficient long-term tracking with local-global similar object interference suppression高效长期跟踪与局部-全局相似目标干扰抑制
PREAI
分享收藏
查看更多内容