IamBatman07/Diagram-Structure-Extractor

Diagram Structure Extractor

Upload an architecture diagram and get its structure back as schema-valid JSON: every text label, box, arrow, icon, and the relationships between components. Output is a typed Pydantic model, so the JSON parses on every call.

Endpoints

• —POST /extract — multipart image upload; returns the typed structure, an annotated PNG, and a flat edge CSV
• —GET /health — liveness

Measured results (15 public diagrams with hand-written ground truth)

• —Schema-valid JSON on 15 of 15 inputs, at every ablation stage
• —Text detection F1 0.949 (EasyOCR), boxes 0.808 (Canny + contours), icons 1.000 (template matching)
• —Arrow detection is the weak stage — F1 0.434, recall 0.346 — and relationship quality is capped by it

Source: https://github.com/Mark007-R/Diagram-Structure-Extractor