wsxwj123/opencode-skills-backup/hivision-idphotos

Hivision ID Photo Processor

Overview

Task skill for executing HivisionIDPhotos image workflows from terminal commands. This skill is for real processing outputs, not repository analysis.

When to Use

Use this skill when user requests include:

• Modify ID photo / 修改证件照
• Cutout / 抠图
• Change background color / 换底色
• Replace background / 换背景
• Crop to standard size / 裁剪
• Generate print layout / 排版
• Remove background / portrait matting
• Change ID photo background color
• Crop to passport or visa size
• Generate print layout sheets (6-inch / 5-inch / A4 style by size settings)
• End-to-end pipeline from raw portrait to final files

Do not use this skill for generic Photoshop editing or non-ID-photo composition work.

Trigger Examples (Natural Language)

The following user phrases should trigger this skill:

• 修改证件照
• 证件照抠图
• 证件照换底色
• 证件照换背景
• 证件照裁剪成1寸
• 证件照排版
• 把这张照片做成证件照

Required Inputs

• input image path
• outdir output directory
• width and height in pixels (default 295x413)
• optional bg_color in hex without # (default 638cce)

Command Entry

Run the wrapper script:

python3 scripts/hivision_task.py <command> [options]

Commands:

• pipeline: matting -> crop -> background -> layout
• matting: output transparent png
• crop: crop transparent id photo by size
• background: apply pure/gradient background to RGBA image
• layout: generate print layout from final ID photo

Typical Usage

python3 scripts/hivision_task.py pipeline \
  --input /abs/path/photo.jpg \
  --outdir /abs/path/out \
  --width 295 --height 413 \
  --bg-color 438edb \
  --render 0

Expected outputs in outdir:

• idphoto_matting.png
• idphoto_crop.png
• idphoto_crop_hd.png
• idphoto_bg.jpg
• idphoto_layout.jpg

Environment Notes

• By default wrapper reads repo path from HIVISION_REPO.
• Default repo path: /Users/wsxwj/.codex/skills/hivision-idphotos/vendor/HivisionIDPhotos
• Ensure model weights exist under:
• hivision/creator/weights/
• (optional) hivision/creator/retinaface/weights/

If model files are missing, download them with the upstream repo script: python3 scripts/download_model.py --models hivision_modnet