Deep Reinforcement Learning for Visual Object Tracking in Videos
lastmod 2026-11-10created 2017-03-022 edits
Deep Reinforcement Learning for Visual Object Tracking in Videos is a paper that treats single-object tracking in video as a sequential decision-making problem and solves it with deep reinforcement learning. It extracts frame features with a convolutional neural network, carries temporal information forward with a recurrent neural network, and trains the tracker to output the object's location as an action at every frame. This page archives the original arXiv paper.
Documentation
- Deep Reinforcement Learning for Visual Object Tracking in Videos
- https://arxiv.org/abs/1701.08936v1
-
1701.08936v1.pdf