Skip to content

Navifra-Kerry/SemanticSegmentation-Libtorch

Repository files navigation

LibTorch Example of SemanticSegmentation

https://blog.kerrycho.com/Image-Segmentation-Libtorch/

GitHub Actions status

Current Support CMake

  • Windows
  • Linux
  • MacOS
windows
mkdir build & cd build
cmake ..

linux
bash build.sh

Example

#Receive data through commandline.
1. Open cmd.exe

2. training or inference
#example training 
D:\vision>SemanticSegmentation-Libtorch.exe train data_dir 

#example inference
D:\vision>SemanticSegmentation-Libtorch.exe inference image_path model_path 

DataSet

# Only Coco data sets are supported.

diretory_root
 cocodataset
  -annotations
   -instances_train2017.json
   -instances_tran2017.json
   -...
  -train2017
   - train images
  -val2017
   - val images

Requirements

  • OpenCV 4.1.1
  • Libtorch 1.4
  • CUDA 10.2
  • Visual studio 2019
  • Windows 10

Install with Scripts

SemanticSegmentation-Libtorch
 - Scripts
    -install.bat
    
Run install.bat
It will download and install automatically.

GitHub Actions status

Convert Python Model to C++

support Only-Resenet backbones
convert.py
Run Python Script

test

GitHub Actions status