This notebook aim to gather the contributions of both group working on deep learning approach to visually track an object in a video. Two different framework were used : GOTURN and YOLO.
The architecture used by GOTURN algorithm is described in the original paper :
Learning to Track at 100 FPS with Deep Regression Networks,
David Held,
Sebastian Thrun,
Silvio Savarese,
The authors has implemented it with caffe : davheld/GOTURN
The tensorflow implementation that has been used to develop our own code: tangyuhao/GOTURN-Tensorflow
Illustration of how this network works:
- Link to slides : Slides
-
GOTURN implemented in a colab ipython Notebook can be found in the following : Code
-
YOLO Implemented in C : Chay16/TrackingVideo