Commit Graph

4 Commits

Author SHA1 Message Date
Adriano 1cc7881a51 feat: pm2d.eval - validation harness CLI per LineShapeMatcher
Tool da CLI per misurare oggettivamente la qualita' del matcher
su dataset etichettato. Halcon ha questo solo nell'IDE (HDevelop),
qui esposto come modulo Python testabile in CI.

Format dataset JSON:
  - template + mask
  - params init matcher (override)
  - find_params (override per find())
  - scenes con ground_truth: lista pose attese (cx, cy, angle, scale,
    tolerance_px, tolerance_deg)

Metriche per scena: TP/FP/FN, precision, recall, IoU medio bbox,
tempo find. Aggregato: precision globale, recall, F1.

Match-to-GT criterio: distanza centro <= tolerance_px AND
|angle| <= tolerance_deg, oppure IoU bbox >= 0.3.

Use case:
- regressione: confronto config A vs B oggettivo
- tuning: trovare param ottimi via grid-search guidato da F1
- validazione pre-deploy: report TP/FP/FN su dataset prod

Esposto come entry-point pm2d-eval (pyproject.toml).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-05 10:09:45 +02:00
Adriano fd7585acc5 feat: interfaccia web HTML (FastAPI + canvas JS)
Sostituisce GUI cv2/tkinter con webapp standalone:

Server (pm2d/web/server.py):
- FastAPI + uvicorn
- Endpoint: GET /, POST /upload, POST /match, POST /auto_tune,
  GET /image/{id}/raw
- In-memory image store (uuid-based)
- Rendering annotated server-side via opencv (overlay bbox + edges
  template warpati)

Frontend (pm2d/web/static/):
- index.html: layout 3 colonne (MODELLO | SCENA | PARAMETRI) + footer
  legenda
- style.css: tema dark, CSS grid responsive
- app.js: canvas HTML5 per visualizzazione scalata fit,
  ROI selection con drag mouse, form parametri live,
  MATCH button, Auto-tune button

Parametri modificabili INLINE (niente dialog separata).
Enter su qualsiasi campo triggera MATCH.
Legenda match in fondo con pallino colorato + dati.

main.py ora lancia il server webapp. Deprecato ingresso GUI cv2
(pm2d/gui.py resta importable per backward compat).

Test: /match su rings_and_nuts: 3/3 ruote in 1.14s (train 0.36s + find 0.77s).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-24 09:06:25 +02:00
Adriano b20b11c029 perf: Numba JIT kernel per score_by_shift (2.1x speedup)
- Nuovo modulo pm2d/_jit_kernels.py con _jit_score_by_shift Numba njit
  parallel + fastmath + boundscheck=False
- Parallelizzazione per riga output (no race condition su acc)
- Fallback automatico numpy se numba non installato
- Warmup automatico al module import (evita JIT lag al 1 match)

Benchmark clip.png (13 istanze):
  prima (numpy + threads): 1.55s
  dopo (numba + threads):  0.72s
  speedup: 2.1x

Pipeline totale full (refine+subpix): 0.80s

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-24 01:30:31 +02:00
Adriano b9a4d51fac feat: PM2D standalone shape-based matcher
Programma standalone Pattern Matching 2D con GUI cv2/tk + algoritmo
puro riusabile. Due backend:

- LineShapeMatcher (default): porting Python di line2Dup (linemod-style)
  - Gradient orientation quantized 8-bin modulo π + spreading
  - Feature sparse top-magnitude con spacing minimo
  - Score via shift-add vettorizzato numpy (O(N_features·H·W))
  - Piramide multi-risoluzione con pruning varianti al top-level
  - Supporto mask binaria per modello non-rettangolare

- EdgeShapeMatcher (fallback): Canny + matchTemplate multi-rotazione

GUI separata da algoritmo. Benchmark clip.png (13 istanze):
  - Edge backend:  84s, 6/13 score ~0.3
  - Line backend:  4.1s, 13/13 score 0.98-1.00

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-24 00:46:59 +02:00