README.md

metadata

title: Floor Plan Segmentation
emoji: 🏠
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 5.23.0
app_file: app.py
pinned: false
license: mit

Floor Plan Segmentation API

Segments rooms, walls, doors, and windows from floor plan images.

Models

• Primary: Mask2Former instance segmentation (pre-trained on floor plans)
• Fallback: OpenCV contour-based detection (no GPU needed)

API Usage

import requests

# Via Gradio API
response = requests.post(
    "https://<space-name>.hf.space/api/predict",
    json={"data": ["<base64_image>"]}
)
result = response.json()

Output Format

All coordinates are normalized (0.0 to 1.0):

{
  "rooms": [{"name": "Wohnzimmer", "type": "living_room", "boundary": [{"x": 0.1, "y": 0.2}, ...], "confidence": 0.85}],
  "walls": [{"start": {"x": 0.0, "y": 0.1}, "end": {"x": 0.5, "y": 0.1}, "is_exterior": true}],
  "doors": [{"position": {"x": 0.3, "y": 0.5}, "width_mm": 900, "type": "standard"}],
  "windows": [],
  "_coordSystem": "normalized",
  "_version": "hf-space-v1"
}