HivisionIDPhotos is a lightweight, efficient AI-powered tool for creating standard ID photos. It targets users needing quick, high-quality ID photos, offering offline processing and customizable output formats.

How It Works

The tool utilizes a pipeline of AI models for tasks like background removal (matting) and face detection. It supports multiple matting models (MODNet, hivision_modnet, rmbg-1.4, birefnet-v1-lite) and face detection models (MTCNN, RetinaFace, Face++ API), allowing users to balance speed and accuracy. The system can generate standard ID photos, print-ready layouts, and supports features like background color changes and beautification.

Quick Start & Requirements

• Install: Clone the repository and install dependencies via pip install -r requirements.txt and pip install -r requirements-app.txt.
• Models: Download model weights using python scripts/download_model.py --models all or manually.
• Run: Launch the Gradio demo with python app.py.
• Prerequisites: Python >= 3.7 (tested on 3.10), Linux, Windows, macOS. Optional GPU acceleration requires CUDA and cuDNN.
• Docs: README_EN.md, API Documentation, Gradio Demo.

Highlighted Details

• Lightweight, CPU-first inference for matting.
• Supports multiple AI models for matting and face detection.
• Generates standard ID photos and print layouts (e.g., 6-inch photo paper).
• Offers an API service and Docker deployment.
• Gradio demo includes features like print layout options, custom background colors, and face alignment.

Maintenance & Community

The project is actively maintained with frequent updates. A vibrant community contributes extensions like ComfyUI workflows, WeChat mini-programs, and a C++ version. Links to community projects and contact information are provided.

Licensing & Compatibility

Licensed under the Apache-2.0 License, permitting commercial use and integration with closed-source projects.

Limitations & Caveats

GPU acceleration is primarily supported for the birefnet-v1-lite model and requires significant VRAM (around 16GB). Some advanced features like "smart suit changing" are marked as "waiting." CPU inference times can vary significantly depending on the chosen models.