hamolato/YOLOE_visual_prompt_demo

YOLOE Visual Prompt Segmentation

This Hugging Face Space demonstrates YOLOE visual prompt segmentation using a reference image and box prompts.

Features

• —Upload a reference image (Image A).
• —Draw bounding boxes directly on Image A.
• —Use these boxes as visual prompts for YOLOE.
• —Upload a target image (Image B).
• —Run visual-prompt-based segmentation / detection inference.

Tech Stack

• —Docker Space
• —Python 3.11
• —Gradio 4.44.1
• —gradio-box-promptable-image
• —Ultralytics YOLOE
• —OpenCV
• —NumPy

Files

• —app.py: Main app.
• —requirements.txt: Python dependencies.
• —Dockerfile: Runtime definition for the Docker Space.
• —.dockerignore: Optional file to reduce build context.

Workflow

1. 1.Upload Image A.
2. 2.Draw one or more bounding boxes.
3. 3.Upload Image B.
4. 4.Run inference.
5. 5.View the segmentation / detection result.

Notes

• —This version focuses on stable box-prompt input.
• —The prompt UI is based on gradio-box-promptable-image, which requires Gradio 4.x.
• —Docker Space is used so the environment can be fully controlled.