 xiaden
|
52a06d6e53
|
Split describe into OCR node + 6 dedicated tagger nodes with full knobs
- NEW tagging/ subpackage with 6 one-tagger-per-node:
ImgUtilsWD14, ImgUtilsDeepDanbooru, ImgUtilsDeepGelbooru,
ImgUtilsMLDanbooru, ImgUtilsCamie, ImgUtilsPixAI
- NEW ImgUtilsOCR: PaddleOCR text extraction with heat/box threshold knobs
- ImgUtilsDetect: added detect_text_with_ocr (bbox detection)
- Removed ImgUtilsDescribe (split into OCR + taggers)
- Each tagger exposes its native threshold/flag controls
|
2026-07-03 01:02:22 +00:00 |
|
|
xiaden
|
91b6c7ec6c
|
Split describe into describe+detect, add confidence knob, clean outputs
- ImgUtilsValidate: outputs now (label, score, full_response) — label is text-only
- ImgUtilsDescribe: tagging + OCR only, dual outputs (tags names, tags+scores)
- ImgUtilsDetect: NEW — face/hand/head/eye/person/nudenet/censor/booru detection
with confidence threshold knob, dual outputs (detections summary, boxes detail)
- ImgUtilsTags: registered in __init__.py
|
2026-07-02 23:14:28 +00:00 |
|