Xiu-Shen WEI （魏秀参）

Xiu-Shen WEI
Ph.D., Research Director
Megvii Research Nanjing (旷视南京研究院)

Office: 15th Floor, Tower A, No. 6 Xingzhi Rd., Nanjing
Email: weixs.gmgmail.com or weixiushenmegvii.com

Media:

About Me

I am currently the Research Director of Megvii Research Nanjing, Megvii Technology. We focus on developing novel computer vision systems, creating new deep learning models, publishing high-quality papers and deploying cutting-edge technologies to better serve Megvii's business.

I got my Ph.D. in LAMDA Group at Nanjing University, advised by Prof. Jianxin Wu. Before that, I received my B.Sc. degree in Computer Science and Technology in 2012, and got my M.Sc. degree in 2014 under the supervision of Prof. Zhi-Hua Zhou.

From Oct. 2016 to Oct. 2017, I was a visiting scholar in Prof. Chunhua Shen’s group at The University of Adelaide, Australia.

Feb. 27, 2020: Two papers accepted for CVPR 2020: one for long-tailed recognition and one for domain adaptive object detection.
Feb. 20, 2020: Co-organized a workshop on Robust Vision Challenge at ECCV 2020.
Oct. 30, 2019: One paper accepted for IEEE Transactions on Knowledge and Data Engineering.
Jul. 6, 2019: Opened the “Awesome Fine-Grained Image Analysis – Papers, Codes and Datasets” homepage.
Jun. 12, 2019: Our team won the first place in both iNaturalist and Herbarium tracks of FGVC 2019 global competition!
Nov. 1, 2018: Published a Chinese book on convolutional neural networks and computer vision.

My research interests include some sub-fields of Computer Vision and Machine Learning:

Deep Convolutional Neural Networks (DCNN) is a type of feed-forward artificial neural network where the individual neurons are tiled in such a way that they respond to overlapping regions in the visual field, which is widely used in image and video related tasks.

Fine-Grained Image Analysis (FGIA) is a hot topic in computer vision and pattern recognition. The goals of FGIA are localizing the fine-grained objects, recognizing the objects’ subordinate categories, retrieving the fine-grained objects and so on.

Long-Tailed Distribution Learning (LTDL) deals with real-world datasets displaying skewed distributions with a long tail, i.e., a few classes (a.k.a. head classes) occupy most of the data, while most classes (a.k.a. tail classes) have rarely few samples.

General Object Detection is a computer technology related to computer vision and image processing that deals with detecting instances of semantic objects of a certain class in digital images and videos.

Weakly Supervised Learning (WSL), especially Multi-Instance Learning (MIL), is a variation on supervised learning. Instead of receiving a set of instances which are individually labeled, the learner receives a set of labeled bags, each containing many instances.

Bag-of-Words Model (BoW) can be applied to image related tasks by encoding local visual descriptors of one image into a high dimensional vector. In BoW, there also include Vector of the Locally Aggregated Descriptors (VLAD) and Fisher Vector (FV).

My papers are available for download in this Publications page, and here is my Google Scholar Citations profile.

CVPR 2017, 2018, 2019, 2020, ICCV 2017, 2019, ECCV 2018, NIPS 2016, 2018, IJCAI 2018, 2019, 2020, AAAI 2016, 2017, 2018, 2019, 2020, ACCV 2018, BMVC 2019, PRCV 2019.