谭杨 Yang Tan

Welcome! I obtained my PhD degree at Tsinghua-Berkeley Shenzhen Institute (TBSI), Tsinghua University (清华大学), advised by Prof. Yang Li. I also worked closely with Prof. Xiao-Ping Zhang, Prof. Lu Fang, and Prof. Shao-Lun Huang.

Previously I obtained my Master degree at Sun Yat-sen University (中山大学) in 2019 and my Bachelor degree at Xidian University (西安电子科技大学) in 2017.

E-mail: tanyang1231@163.com

GitHub / Google Scholar /

NEWS

[2024-04-07] Our team won the 1st place in ICDAR'24 Competition on Multi-Font Recognition and OCR "Data Alchemist" track.

[2024-01-16] One paper is accepted by IEEE TNNLS (IF=14.225, CAS SCI Q1).

[2023-11-06] One paper is accepted by BIBM'23 Workshop on Deep Learning.

[2023-06-22] One paper is accepted by ICIP'23.

Selected Publications

My research interests include transfer learning, few-shot learning and computer vision. Please see my Google Scholar page for the full publication list.

	Transferability-Guided Cross-Domain Cross-Task Transfer Learning Yang Tan, Enming Zhang, Yang Li, Shao-Lun Huang, Xiao-Ping Zhang IEEE Transactions on Neural Networks and Learning Systems (TNNLS), 2024 link / We remove the requirement on auxiliary tasks of OTCE to make it more efficient and generalizable in practical scenarios. Moreover, by taking OTCE as a loss function, we further improve the transfer performance of the source model.
	Efficient Prediction of Model Transferability in Semantic Segmentation Tasks Yang Tan, Yicong Li, Yang Li, Xiao-Ping Zhang International Conference on Image Processing (ICIP), 2023 We propose a flexible adaptation method to gneralize existing transferability metrics to semantic segmentation tasks. In addition, we propose to utilize pixel-wise transferability map to further enhance model finetuning.
	Finding the Most Transferable Tasks for Brain Image Segmentation Yicong Li, Yang Tan, Jingyun Yang, Yang Li, Xiao-Ping Zhang International Conference on Bioinformatics and Biomedicine (BIBM), 2022 link / arxiv / We propose a framework for finding the most transferable source tasks in Brain Image Segmentation
	OTCE: A Transferability Metric for Cross-Domain Cross-Task Representations Yang Tan, Yang Li, Shao-Lun Huang IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021 (Oral Presentation) link / arxiv / code / We propose an analytical transferability metric to predict the transfer performance of source model (task) under the challenging cross-domain cross-task transfer settings.
	CrossNet++: Cross-Scale Large-Parallax Warping for Reference-Based Super-Resolution Yang Tan, Haitian Zheng, Yinheng Zhu, Xiaoyun Yuan, Xing Lin, David Brady, Lu Fang IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2020 link / We propose an end-to-end framework to fuse multi-scale images under large parallax for mutli-camera systems.
	Justlookup: One millisecond deep feature extraction for point clouds by lookup tables Hongxin Lin, Zelin Xiao, Yang Tan, Hongyang Chao, Shengyong Ding International Conference on Multimedia and Expo (ICME), 2019 (Oral Presentation) link / arxiv / We propose to apply a classical lookup table to speed up the inference process for a particular deep architecture for 3D point cloud tasks.
	Face Recognition from Sequential Sparse 3D Data via Deep Registration Yang Tan, Hongxin Lin, Zelin Xiao, Shengyong Ding, Hongyang Chao International Conference on Biometrics (ICB), 2019 (Oral Presentation) link / arxiv / We propose a method to fuse a sequence of low-quality 3D facial data acquired by portable DoE based structured light system to achieve high recognition accuracy.