Target-Aware Deep Tracking

Pytorch implementation of the Target-Aware Deep Tracking (TADT) method.

Main contents:

• Codes of the TADT tracker.
• Codes of visualization.

Performance

tracker	OTB-50	OTB2013	OTB-100(OTB2015)
TADT-python	0.615	---	0.656
TADT-official	---	0.680	0.660

rate: 77FPS (i7 8700k, RTX2080)

Note: We think that the tiny performance gap between TADT-python and TADT-official is caused by the difference between Matconvnet and pytorch

Environment

This code has been tested on Ubuntu 16.04, Python 3.7, Pytorch 1.1, CUDA 10, RTX 2080 GPU

Requirements

numpy, cv2, matplotlib, scipy, yacs

Installation

1. Clone the GIT repository:
   $ git clone
2. Run the demo script to test the tracker:
   python demo_tadt.py

Contact

Zikun Zhou Email: [email protected]