This is the implementation of CVPR'19 paper "SPM-Tracker: Series-Parallel Matching for Real-Time Visual Object Tracking"
Перейти к файлу
Chong Luo f3e997ce0d first release 2019-08-30 16:13:05 +08:00
configs/spm_tracker first release 2019-08-30 16:13:05 +08:00
data first release 2019-08-30 16:13:05 +08:00
siam_tracker first release 2019-08-30 16:13:05 +08:00
tools first release 2019-08-30 16:13:05 +08:00
.gitignore Initial commit 2019-08-30 01:05:45 -07:00
CODE_OF_CONDUCT.md Initial commit 2019-08-30 01:06:00 -07:00
LICENSE Initial commit 2019-08-30 01:06:03 -07:00
NOTICE.txt first release 2019-08-30 16:13:05 +08:00
README.md first release 2019-08-30 16:13:05 +08:00
SECURITY.md Initial commit 2019-08-30 01:06:01 -07:00
compile.sh first release 2019-08-30 16:13:05 +08:00
requirements.txt first release 2019-08-30 16:13:05 +08:00

README.md

SPM-Tracker

This is an inference-only implementation of SPM-Tracker

  • SPM-Tracker: Series-Parallel Matching for Real-Time Visual Object Tracking, CVPR 2019

Installation

Prerequisite Lib

  • Python3 is required for this project.

  • You should install PyTorch and Torchvision lib firstly. The installation can refer to official repo https://pytorch.org/ .

  • The other Python dependences can be installed by pip.

pip install -r requirements.txt

Compile Cython Extensions

Some components are written in Cython & CUDA. It's necessary to manually compile them.

bash compile.sh

Known issue for torch >= 1.0.0 on Win32 platform: the MSCV compiler cannot handle #warning, see this pull request: https://github.com/pytorch/pytorch/pull/20484


Demo

  • tools/track_video.py shows an example to track a video

You may just simply call

python tools/track_video.py --video data/demo/boy.avi --cfg configs/spm_tracker/alexnet_c42_otb.yaml