YOLO Dataset Auto-Annotate Workflow

세부 정보

파일 다운로드 (1)

이 버전에 대해

This version adds the ability to select which bbox to use in the situation that Florence2Run returns more than one. Typically, I have found that the higher the bbox index, the smaller the area and the more tightly matching the description. Thus, I wanted a way to be able to pick the tightest match, i.e. the bbox with the smallest area. After learning about For Loops, I've added functionality to allow you, the user, to specify that.

모델 설명

This workflow is intended to provide a head start for anyone needing to annotate image sets for yolo custom model training. It processes your image files using Florence2 for initial bbox detection and then SAM2 for finer bbox detection. It is intended to save files in a "/yolo/" sub-directory relative to the image base directory specified. It saves padded 640x640 images (default yolo dimensions) in a "/images" sub-directory and the labels which include the bbox dimensions in yolo format in the "/labels" sub-directory.

If you build on this, please hit me up!

- BillyB

이 모델로 만든 이미지

정렬

모델 유형	워크플로우
기본 모델	Other
게시일	2025-12-27