Image feature extraction method based on a self-supervised triple neural network 基于自监督三重神经网络的图像特征提取方法
National invention patent · inventors Jie Jiang, Tianjian Zhou (2nd inventor), Yishan Li, et al. 国家发明专利 · 主要发明人：蒋捷、周天健（第2发明人）、李易珊 等
Service学术服务
Reviewing, patents, and open science. 审稿、专利与开放科学。
Academic service is part of the job: reading other people's work carefully, and making my own work easy to check and reuse.
学术服务是研究工作的一部分：认真阅读他人的工作，也让自己工作可被检验与复用。
Peer review同行评审
Reviewing.审稿。
I serve as a reviewer for the following conferences and journals.
我担任以下会议与期刊的审稿人。
Conferences会议
- NeurIPS Conference on Neural Information Processing Systems 神经信息处理系统大会
- ICLR International Conference on Learning Representations 国际学习表征会议
- ICML International Conference on Machine Learning 国际机器学习大会
- AAAI AAAI Conference on Artificial Intelligence AAAI 人工智能大会
- CVPR IEEE/CVF Conference on Computer Vision and Pattern Recognition IEEE/CVF 计算机视觉与模式识别会议
Journals期刊
- IEEE TNNLS IEEE Transactions on Neural Networks and Learning Systems 神经网络与学习系统汇刊
- Pattern Recognition Elsevier Elsevier
- Neural Networks Elsevier Elsevier
Multi-scale semantic information enhancement method and apparatus for improving the performance of contrastive learning models 提升对比学习模型性能的多尺度语义信息增强方法和装置
National invention patent · inventors Jie Jiang, Tianjian Zhou (2nd inventor), Yishan Li, et al. 国家发明专利 · 主要发明人：蒋捷、周天健（第2发明人）、李易珊 等
Pedestrian trajectory prediction method, apparatus and device 一种行人轨迹预测方法、装置及设备
National invention patent 国家发明专利
Retinal image segmentation method and apparatus 一种视网膜图像分割方法及装置
National invention patent 国家发明专利
Open science开放科学
Making the work checkable.让工作可被检验。
Visual SSL DigestVisual SSL Digest
I curate the Visual SSL Digest, a daily digest that tracks new work on self-supervised visual representation learning, visual foundation models and geometric intelligence, with a short structured note on each paper. It is public and has no paywall.
我维护《Visual SSL Digest》，每日跟踪自监督视觉表征学习、视觉基础模型与几何智能方面的新工作，并为每篇论文写一段结构化的短评。内容公开，不设付费墙。
Code & project pages代码与项目主页
Code for EquiSSL is released together with its project page, which collects the method, figures, results and BibTeX; DAGA's implementation is released on GitHub. I would rather publish a small, reproducible artefact than a large, unrunnable one.
EquiSSL 的代码与其项目主页一同发布，后者汇总了方法、插图、结果与 BibTeX；DAGA 的实现也已在 GitHub 开源。比起庞大而难以运行的产物，我更愿意发布小而可复现的成果。
Contact联系
Get in touch.欢迎联系。
For reviewing matters, research questions or collaborations, email is the most reliable way to reach me.
关于审稿事务、研究问题或合作，邮件是最可靠的联系方式。
AI for Science & Visual Intelligence Group · College of Systems Engineering · National University of Defense Technology · Changsha 410073, China 科学智能与视觉智能研究组 · 系统工程学院 · 国防科技大学 · 中国长沙 410073