Awesome

Residual-learning-based-two-stream-network-for-RGB-T-object-tracking

This repository contains the codes for paper "Residual learning-based two-stream network for RGB-T object tracking" by Yili Chen, Minjie Wan*, Yunkai Xu, et al. (*Corresponding author).

The overall repository style is partially borrowed from MANet (https://github.com/Alexadlu/MANet). Thanks to Chenglong Li.

The relative datasets and evaluation toolkits can be downloded from the following links:

RGBT234 dataset

Link：https://pan.baidu.com/s/1weaiBh0_yH2BQni5eTxHgg code：qvsq

RGBT210 dataset

Link：https://pan.baidu.com/s/1FClmX0SH3WarcczkEQbmwA code：ps8j

GTOT dataset

Link：https://pan.baidu.com/s/1zaR6aXh9PVQs063Q_b9zQg code：ajma

RGBT234 toolkit

Link：https://pan.baidu.com/s/1UksOGtD2yl6k8mtB-Wr39A code：4f68

RGBT210 toolkit

Link：https://pan.baidu.com/s/1KHMlbhu5R29CJvundGL4Sw code：8wtc

GTOT toolkit

Link：https://pan.baidu.com/s/1iVVAXS4LZLvoQSGQnz7ROw code：d53m

Requierments:

python2.7
pytorch == 0.3.1
numpy
PIL
by yourself need install some library functions

Usage

Generate train data using .../pretrain/prepro_data.py.
Train your own model using .../pretrain/train.py, the relative parameters can be adjusted in option.py and train.py.
Track with the trained model by running .../tracking/run_tracker.py, some parameters need to be set in .../tracking/options.py and track.py.