# unet_mxnet2ncnn
Train mxnet unet and run it on ncnn.
How to use:
1. Get dataset and put it in the some directory. Adjust files under mxnet-unet/data based on dataset location.
2. cd mxnet-unet; python3 trainunet.py
3. Get inference model by running: python3 train2infer.py
4. Use ncnn tool 'mxnet-ncnn' to get ncnn model
Need to patch tools/mxnet/mxnet2ncnn.cpp for ncnn. Set num_group to 1 in if for "Deconvolution" op. But I think it will fix also if designate it in mxnet network.
5. Run ncnn-unet.
Install OpenCV first.
cd ncnn-unet
Edit CMakefiles to fit your enviroment.
mkdir build
cmake ..; make
./unet your_jpg_file
算法部署_使用NCNN部署MXnet-Unet图像分割算法_优质算法部署项目实战.zip （12个子文件） 
算法部署_使用NCNN部署MXnet-Unet图像分割算法_优质算法部署项目实战 ncnn-unet 
CMakeLists.txt 558B src 
main.cpp 5KB models convert.sh 222B mxnet-unet trainunet.py 4KB unetdataiter.py 5KB data seg_train.txt 3.25MB seg_test.txt 429KB predict.py 2KB unet.py 7KB check.py 3KB train2infer.py 1KB README.md 664B资源评论
