Visual Object Tracking
The task object tracking is vital in numerous applications, such as autonomous driving, intelligent surveillance, robotics, etc. The standard formulation for the object tracking is assigning a so called bounding box to an object in a video stream (i.e., on every frame of the video), provided only the bounding box for a chosen object on the first frame. In this field, we want to achieve not only accurate localization, but also a small number of object losses during tracking.
At the seminar, we will look at the existing state-of-the-art models, discuss the approaches and techniques that are used in this area to achieve the maximum tracking perfomance.
Li, Bo and Yan, Junjie and Wu, Wei and Zhu, Zheng and Hu, Xiaolin: High performance visual tracking with siamese region proposal network
Bai, Shuai and He, Zhiqun and Xu, Ting-Bing and Zhu, Zheng and Dong, Yuan and Bai, Hongliang: Multi-hierarchical Independent Correlation Filters for Visual Tracking
Speaker: Vladislav Belyaev.
Presentation language: Russian.
Date and time: April 3rd, 18:30-20:00.
Location: Times, room 204.
Videos from previous seminars are available at http://bit.ly/MLJBSeminars
- About seminars
29 May 2019ICLR 2019 Overview
22 May 2019Segmentation in 2019. The fastest and the most accurate.
8 May 2019Speech recognition and speech synthesis
17 April 2019Open Questions about Generative Adversarial Networks
10 April 2019Adaptive Sampled Softmax with Kernel Based Sampling