vietanhdev/anylabeling

Is it possible to create a SAM mask after inference directly from a custom yolov5 model?

Opened this issue · 1 comments

YoungjaeDev commented

I wonder answer about my question

  1. Or is it possible to load a pre-predicted bbox?
  2. I wonder if it is possible to perform the task by dividing it into instance seg and semantic seg tasks.
ryouchinsa commented

You can export SAM 2.1 ONNX models for image prediction using export_onnx.py in sam-cpp-macos repository.

After detecting bounding boxes using YOLO11, you can segment inside the box using Segment Anything Model2.

boxes2polygons.mp4