Pyon2-Detector

Scripts to train "Anime face detector" model using deep learning, and detect anime faces with the trained model.
For more details, please refer to this page.

Prepare Dataset

place dataset like this.

  dataset
   /train
     /0      -- 0 labeled dataset
     /1      -- 1 labeled dataset
     ...
   /valid
     /0      -- 0 labeled dataset
     /1      -- 1 labeled dataset
     ...

Usage

# resize images
./pyon2-detector.py gendata --size=[output image size] --rawdata=[dataset directory] --output=[resized dataset directory]

# generate mean file
./pyon2-detector.py genmean --train=[training dataset directory] --output=[mean file path]

# generate train data list and validation data list
./pyon2-detector.py genlist --dataset=[training data directory] --output=[training data list]
./pyon2-detector.py genlist --dataset=[validation data directory] --output=[validation data list]

# train model
./pyon2-detector.py learn --train=[training data list] --valid=[validation data list] --gpu=[gpu flag] --mean=[mean file path] --model=[model name]

# detect faces
./pyon2-detector.py detect --gpu=[gpu flag] --mean=[mean file path] --model=[model dump file path] --input=[input image path] --output=[output image path]

Reqirements

1. chainer >= 1.3.0
2. OpenCV >= 3.0