LightTrack (inference based on openvino)

官方 pytorch 代码仓库：

https://github.com/researchmm/LightTrack

论文：

LightTrack: Finding Lightweight Neural Networks for Object Tracking via One-Shot Architecture Search

介绍

本仓库基于 Intel OpenVINO Toolkit 部署 LightTrack 跟踪算法，包含 Python、C++ 两种语言的推理代码 。

优势：方便部署，高性能。

本仓库的推理模型将预处理和部分后处理融入模型之中，使部署代码量更少，更加方便，并且推理引擎使得预处理速度更快。

推理速度

Intel CPU	preprocess+inference+postprocess average time
i7-11700K	3.4ms
i7-10710U	5.5ms
i7-7700HQ	7.5ms

模型修改与导出 tutorials

README

安装 OpenVINO Toolkit

参考官网安装教程 Get Started Guides

wget https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
sudo apt-key add GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
echo "deb https://apt.repos.intel.com/openvino/2022 bionic main" | sudo tee /etc/apt/sources.list.d/intel-openvino-2022.list
sudo apt update
apt-cache search openvino
sudo apt install openvino

Run this command in shell. (Every time before using OpenVINO)

source /opt/intel/openvino_2022/setupvars.sh

安装python(3.8)依赖

cd /opt/intel/openvino_2022/tools
pip install -r requirements[onnx].txt

C++ demo Build and Run

build

cd /your_path/LightTrack_openvino/Cpp_Infer
mkdir build && cd build
cmake .. && make -j

run

视频文件输入:

./LightTrack 0 "../../images/bag.avi"

摄像头输入:

./LightTrack 1 0

图片序列输入:

./LightTrack 2 "../../images/Woman/img/%04d.jpg"

Python demo Run

视频文件输入:

python infer.py --mode 0 --video "../images/bag.avi"

摄像头输入:

python infer.py --mode 1

图片序列输入:

python infer.py --mode 2 --image_path "../images/Woman/img/*.jpg"