subinium / anime-face-detector

A Faster-RCNN based anime face detector implementation using tensorflow.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool


A Faster-RCNN based anime face detector.

This detector is trained on 6000 training samples and 641 testing samples, randomly selected from the dataset which is crawled from top 100 pixiv daily ranking.

Thanks to OpenCV based Anime face detector written by nagadomi, which helps labelling the data.

The original implementation of Faster-RCNN using Tensorflow can be found here


  • Python >= 3.6
  • tensorflow latest 1.x or 2.x
  • opencv-python (Will use other packages like pillow and scikit-image as backend in future version)
  • cython (optional, can be ignored with additional -nms-type PY_NMS argument)
  • Pre-trained ResNet101 model


  1. Clone this repository
    git clone
  2. Download the pre-trained model
    Google Drive: here
    Baidu Netdisk: here
  3. Unzip the model file into model directory
  4. Build the CPU NMS model (skip this step if use PY_NMS with argument: -nms-type PY_NMS)
    make clean
    If using Windows Power Shell, type cmd /C make.bat to run build script.
  5. Run the demo as you want
    • Visualize the result (without output path):
      python -i /path/to/image.jpg
    • Save results to a json file
      python -i /path/to/image.jpg -o /path/to/output.json
      Format: {"image_path": [{"score": predicted_probability, "bbox": [min_x, min_y, max_x, max_y]}, ...], ...} Sample output file:
      {"/path/to/image.jpg": [{"score": 0.9999708, "bbox": [551.3375, 314.50253, 729.2599, 485.25674]}]}
    • Detecting a whole directory with recursion
      python -i /path/to/dir -o /path/to/output.json
    • Customize threshold
      python -i /path/to/image.jpg -nms 0.3 -conf 0.8
    • Customize model path
      python -i /path/to/image.jpg -model /path/to/model.ckpt
    • Customize nms type (supports CPU_NMS and PY_NMS, not supports GPU_NMS because of the complicated build process for Windows platform)
      python -i /path/to/image.jpg -nms-type PY_NMS
    • Crop detected images and store them in a folder (start output is an integer to start naming the cropped images, default is 0)
      python -i /path/to/image/or/folder -crop-location /path/to/store/cropped/images -start-output 1
    • Crop detected images and resizes them
      python -i /path/to/image/or/folder -crop-location /path/to/store/cropped/images -crop-height 224 -crop-width 224


Mean AP for this model: 0.9086

Copyright info: 東方まとめ by 羽々斬

Copyright info: 【C94】桜と刀 by 幻像黒兎

Copyright info: アイドルマスター シンデレラガールズ by 我美蘭@1日目 東A-40a

About training

This model is directly trained by Faster-RCNN, with following argument:

python tools/ --weight data/imagenet_weights/res101.ckpt --imdb voc_2007_trainval --imdbval voc_2007_test --iters 60000 --cfg experiments/cfgs/res101.yml --net res101 --set ANCHOR_SCALES "[4,8,16,32]" ANCHOR_RATIOS "[1]" TRAIN.STEPSIZE "[50000]"


We've uploaded the dataset to Google drive here, dataset structure is similar to VOC2007 (used in original Faster-RCNN implementation).

Citation and declaration

Feel free to cite this repo and dataset.
This work is not related to my research team and lab, just my personal interest.

ezoic increase your site revenue


A Faster-RCNN based anime face detector implementation using tensorflow.

License:MIT License


Language:Python 97.9%Language:Cuda 1.9%Language:Batchfile 0.1%Language:C++ 0.1%Language:Makefile 0.0%