A classification using mmpretrain trained to classify western images based on ConvNeXtV2-tiny.Used for classifying anime images based on whether they are in the Western style.
The evaluation accuracy on the validation set is 95%.
Trained using 7,000 Western images and 8,000 non-Western images, with the Western training set sampled from e-hentai.
Of course, this model also has many shortcomings, such as a very low recognition accuracy for line-drawing images.
Huggingface space:https://huggingface.co/spaces/TLME/western-anime-images-classification
How to use
Python>=3.9
Install pytorch
pip install -r requirements.txt
edit infer.py , change "path = './testimg/'" to your target folder
python infer.py
- Downloads last month
- 0