metadata
title: Apartment Before-After Object Comparison
emoji: 🏠
colorFrom: indigo
colorTo: purple
sdk: docker
pinned: false
license: mit
🏠 DINO (RT-DETR) + SAM 2/3 + ByteTrack Apartment Inspector
This Hugging Face Space hosts the Apartment Before-After Object Comparison application, built using Gradio, Ultralytics RT-DETR (DINO architecture), ByteTrack/BoT-SORT tracking, and the Segment Anything Model (SAM 2/SAM 3) for pixel-level mask segmentation.
Features
- Dual Video Input: Upload and analyze "Before" and "After" videos of apartment spaces.
- Deep Tracking & Segmentation: Combined class-aware detection (RT-DETR/YOLO) with high-fidelity segmentation (SAM 2/SAM 3).
- Persistent Tracking IDs: Assigns persistent IDs to objects across frames using ByteTrack or BoT-SORT to keep counts precise and handle occlusions.
- Detailed Inventory Reports: Outputs differences between Before and After scenes, showing added/removed counts and detailed lists of specific tracking IDs per category (e.g.
Chair 🪑: 3 items (IDs: #1, #3, #4)). - Interactive Tuning: Control Confidence, IoU, tracking parameters, and frame rates dynamically.
Run Locally
You can run this application locally by installing the dependencies:
pip install -r requirements.txt
python app.py
Or by building the Docker container:
docker build -t apartment-inspector .
docker run -p 7860:7860 apartment-inspector