Skip to content

Darkflow Movidus integration for tiny yolo v2 object detection

Notifications You must be signed in to change notification settings

tiantzp/yolo-darkflow-movidius

 
 

Repository files navigation

Need to have darkflow (https://github.com/thtrieu/darkflow) installed in the path.

You can follow the next steps o run the install.sh script:

git clone https://github.com/fernandodelacalle/yolo-darkflow-movidius.git

cd yolo-darkflow-movidius/
pip install --upgrade cython

git clone https://github.com/thtrieu/darkflow darkflow_all
cd darkflow_all
python3 setup.py build_ext --inplace
pip install -e .

cd ..
ln -s darkflow_all/darkflow darkflow

Next download the tiny yolo v2 weights for voc and save the built graph to a protobuf file (.pb):

cd darkflow_all
wget -O tiny-yolo-voc.cfg https://raw.githubusercontent.com/pjreddie/darknet/master/cfg/yolov2-tiny-voc.cfg
wget -O tiny-yolo-voc.weights https://pjreddie.com/media/files/yolov2-tiny-voc.weights
## Saving graph and weights to protobuf file
python3 flow --model tiny-yolo-voc.cfg --load tiny-yolo-voc.weights --savepb
cd ..
mv  darkflow_all/built_graph/ . 

Finally compile the pb to a movidus graph as follows:

mvNCCompile built_graph/tiny-yolo-voc.pb -s 12 -in input -on output -o built_graph/tiny-yolo-voc.graph

Run yolov2 in the movidius with:

python3 test_movidius.py -i video.avi

About

Darkflow Movidus integration for tiny yolo v2 object detection

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 93.2%
  • Shell 5.1%
  • HTML 1.7%