sarkarritesh2003/clearview-backend

Clearview Backend

FastAPI/Gradio service behind the Clearview app: dehazes/clears a photo, detects objects (YOLOv8n), scores each for visible damage/wear, and returns both a clean enhanced image and an annotated one plus a JSON report.

Endpoints

• —GET /health -- liveness check
• —POST /analyze -- multipart form:
• —image (file, required)
• —threshold (float, default 0.65) -- scuff-score cutoff for "scuffed"
• —warn_ratio (float, default 0.75) -- fraction of threshold for "possible wear"
• —detector (voc or regions, default voc) -- kept for backward compatibility; voc now runs YOLOv8n regardless of the name
• —max_boxes (int, default 25)
• —no_dehaze (bool, default false)
• —no_enhance (bool, default false)
• —sharpen (float, default 0.5)

Returns JSON: { report, original_image_base64, original_image_content_type, modified_image_base64, annotated_image_base64, processing_seconds }

See the main project README for the full pipeline breakdown (dehaze.py, scuffscore.py, enhance.py, detector.py, regiondetector.py).