IndraDThor/retail-shelf-detection

Retail Shelf Detection Pipeline

This Hugging Face Space runs a retail shelf monitoring demo using:

• —CLIP for empty-shelf filtering
• —Grounding DINO for grocery-item detection
• —rule-based stock classification for Empty, Low, and OK
• —a secondary misplaced-product flag

Inputs supported:

• —image upload
• —webcam snapshot
• —video upload

Code structure:

• —app.py: application entry point
• —shelf_monitor/core: typed configuration and logging
• —shelf_monitor/schemas.py: Pydantic models for app data
• —shelf_monitor/services: inference, geometry, annotation, and video analysis
• —tests/: lightweight regression tests for deterministic logic

Author: Suresh Lama Hugging Face: https://huggingface.co/IndraDThor GitHub: https://github.com/Tokol/