jiuxianghedonglu / AnimeHeadDetection

Anime head detection using faster-rcnn-fpn

Anime Head Detection

1. Dataset

All the images in the dataset are collected from konachan.net and labeled by myself with LabelImg. Download dataset

2. Usage

2.1 Requirements

• python 3.6
• pytorch 1.3.0
• torchvision 0.4.1
• opencv-python 3.4.1

2.2 Training

1. Download dataset. Following are dataset directory trees and labeled file format.

    data
        imgs
            anime_0001.jpg
            anime_0002.jpg
            ...
        tr_label.csv
        val_label.csv
        ...

    file_name,xmin ymin xmax ymax label;...;xmin ymin xmax ymax label,img_width,img_height

2. run python train.py

2.3 Detecting image

1. You can download pretrained model weights and put into checkpoints/weights or train it by yourself.
2. run python detect_image.py your_image_path
3. Example:

2.4 Detecting video

1. You can download pretrained model weights and put into checkpoints/weights or train it by yourself.
2. run python detect_video.py your_video_path
3. Example: alt text