This repository provides an anime face detector and landmark predictor for near-frontal anime faces, built upon the MMDetection and MMPose frameworks. It's designed for researchers and developers working with anime-style imagery who need to locate faces and key facial features.

How It Works

The detector leverages MMDetection for object detection and MMPose for landmark prediction. It utilizes pre-trained models from these frameworks, specifically adapted for anime aesthetics. The approach allows for efficient and accurate detection of anime faces and the subsequent prediction of 28 landmark points on each detected face.

Quick Start & Requirements

• Install via pip: pip install anime-face-detector
• Prerequisites: openmim, mmcv-full, mmdet, mmpose.
• Tested on Ubuntu.
• Demo available via Gradio: pip install gradio, git clone https://github.com/hysts/anime-face-detector, cd anime-face-detector, python demo_gradio.py.

Highlighted Details

• Detects near-frontal anime faces.
• Predicts 28 landmark points per face.
• Uses models from MMDetection and MMPose.
• Demo uses images generated by the TADNE model to avoid copyright issues.

Maintenance & Community

• Primary contributor: hysts.
• Links to MMDetection and MMPose repositories are provided.

Licensing & Compatibility

• The repository itself does not explicitly state a license. However, it relies on MMDetection and MMPose, which are typically released under permissive licenses like Apache 2.0. Compatibility for commercial use would depend on the licenses of the underlying frameworks and any specific model weights used.

Limitations & Caveats

• The package is explicitly tested only on Ubuntu.
• The README does not specify the license for the repository itself, which could impact commercial use.