Shiyi Lan

Shiyi Lan 兰石懿

I am a Ph.D. candidate supervised by Dr. Larry S. Davis at University of Maryland, College Park. I obtained my Bachelor's degree of Computer Science at Fudan University. Currently, I'm on the job market.

My research interests lie in the fields of 2D and 3D Object Detection , and Instance Segmentation. My long-term research goal is to apply Continual Learning on Detection and Segmentation, e.g. Weakly/Semi- supervised Instance Segmentation, Weakly/Semi- supervised Object Detection, and Semi-supervised 3D Object Detection.

Email: sylan [at] umd [dot] edu

CV / Google Scholar / GitHub

Publications

	DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision Shiyi Lan, Zhiding Yu, Christopher Choy, Subhashree Radhakrishnan, Guilin Liu, Yuke Zhu, Larry S. Davis, Anima Anandkumar ICCV, 2021 arXiv / code The first-place weakly supervised instance segmentation model using box labels.
	M3DETR: Multi-representation, Multi-scale, Mutual-relation 3D Object Detection with Transformers Tianrui Guan, Jun Wang, Shiyi Lan†, Rohan Chandra, Zuxuan Wu, Larry S. Davis, Dinesh Manocha † Correspondence Author. * means equal contribution. WACV, 2022 arXiv / code The multi-representation, multi-scale, mutual-relation 3D object detector with transformers.
	InfoFocus: 3D Object Detection for Autonomous Driving with Dynamic Information Modeling Jun Wang, Shiyi Lan, Mingfei Gao, Yi Wu, Larry S. Davis, * means equal contribution. ECCV, 2020 arXiv 3D Object Detection with the effective dynamic attention module.
	SaccadeNet: A Fast and Accurate Object Detector Shiyi Lan, Zhou Ren, Yi Wu, Larry S. Davis, Gang Hua CVPR, 2020 arXiv / code A real-time and high-performance object detector.
	Modeling Local Geometric Structure of 3D Point Clouds Using Geo-CNN Shiyi Lan, Ruichi Yu, Gang Yu, Larry S. Davis CVPR, 2019 arXiv / code Modeling point cloud by leveraging geometric information of point clouds.
	FastMask: Segment Multi-scale Object Candidates in One Shot Hexiang Hu, Shiyi Lan, Yuning Jiang, Zhimin Cao, Fei Sha * means equal contribution. CVPR, 2017 (Spotlight) arXiv / code Generating instance segmentation candidates by leveraging multi-scale feature paramids.

Selected Projects

	Neural style transfer on Android. Fudan University, 2017 This project is able to run neural style transfer on Android devices with less 300MB memory. To achieve this goal, I applied the highly optimized deep learning architecture MXNet and extremely compact neural network SqueezeNet.
	ASA.tv Fudan University, 2015 code A full-stack project that implements basic functions of a UGC (User Generated Content) Video Platform. I implemented the backend and frontend independently with Django (Python) and ReactJS (JavaScript).
	Canvas Danmaku Fudan University, 2015 code A full-stack project that implements basic functions of online video player with floating comments. I implemented the backend and frontend independently with KoaJS (NodeJS) and ReactJS (JavaScript).
	Chunked Videos Uploader Fudan University, 2015 code A smart widget for uploading videos written in Python. The users can start/suspend/resume uploading the video. The videos are chunked before uploading for robustness and safety. This widget is designed for the user of ASA.tv to upload their videos.

Experience