| Jishi online sharing issue 98 |
For a long time, in order to let everyone better understand the excellent papers and work in the academic field, Jishi has invited more than 100 technical experts and guests, and completed 97 Jishi online live broadcast sharing. For previous sharing, please go to bbs.cvmart.net/topics/149. Friends are also welcome to recommend themselves or recommend more excellent technical guests to Jishi for technology sharing, and exchange and learn with everyone~~
In this sharing, we invited Yan Bin, a doctoral student at IIAU-LAB, School of Information and Communication, Dalian University of Technology, to introduce their latest work on ECCV2022 Oral:
Towards Grand Unification of Object Tracking
"The field of target tracking can be divided into the following four sub-tasks: single target tracking, video target segmentation, multi-target tracking, multi-target tracking and segmentation. Due to the differences in the definitions and settings of different sub-tasks, most existing tracking Algorithms are designed to solve single or partial sub-tasks, and lack the ability to expand to other tasks. Obviously, their long-term fragmented state is not friendly to the development of the tracking field.
In this sharing, we will introduce Unicorn, a large unified model for target tracking tasks. This model only needs one set of network structure and one set of model parameters to complete four tracking tasks at the same time. This work has been accepted by ECCV2022 as Oral. For the first time, Unicorn has achieved the unification of the target tracking network structure and learning paradigm, and achieved excellent performance on 8 challenging datasets from four tracking tasks, and refreshed the State-of -The-Art. "
Live address: http://live.bilibili.com/3344545
01 Live information
Time: July 26, 2022 (Tuesday) 20:00-21:00
Topic: ECCV22 Oral Unicorn: Towards a Unified Object Tracking
Live address: http://live.bilibili.com/3344545
02 Guest introduction
Yan Bin
Dalian University of Technology IIAU-LAB doctoral student of the School of Information and Communication, supervised by Professor Lu Huchuan and Professor Wang Dong, research direction is video target tracking, currently as the first author published many papers in top computer vision conferences CVPR, ICCV, ECCV, He has won the VOT Challenge several times, and has been an intern at Microsoft Research Asia and ByteDance Artificial Intelligence Laboratory.
Personal homepage: https://masterbin-iiau.github.io/
03 About sharing
➤ Share Outline
1. Development review of object tracking
2. Towards the unification of target tracking
3、Unicorn
4. Experimental results and Demo
➤Thesis
Towards Grand Unification of Object Tracking
Paper address:
https://arxiv.org/abs/2207.07078
Code address:
https://github.com/MasterBin-IIAU/Unicorn