avant codex

test_yolo/README.md

@@ -0,0 +1,37 @@
+# Test YOLO (feuilles/plantes)
+
+## 1) Créer et activer un venv
+
+```bash
+python3 -m venv .venv
+source .venv/bin/activate
+```
+
+## 2) Installer les dépendances
+
+```bash
+pip install --upgrade pip
+pip install ultralyticsplus==0.0.28 ultralytics==8.0.43 opencv-python matplotlib
+```
+
+## 3) Lancer un test
+
+```bash
+python3 test_yolo/test_yolo_leaf.py --image /chemin/vers/ma_plante.jpg
+```
+
+Sorties générées :
+- `test_yolo/output/detections.json`
+- `test_yolo/output/annotated.jpg`
+
+## 4) Paramètres utiles
+
+```bash
+python3 test_yolo/test_yolo_leaf.py \
+  --image /chemin/vers/ma_plante.jpg \
+  --conf 0.25 \
+  --iou 0.45 \
+  --max-det 1000 \
+  --json-out test_yolo/output/detections.json \
+  --image-out test_yolo/output/annotated.jpg
+```

test_yolo/test_yolo/output/detections.json

@@ -0,0 +1,65 @@
+{
+  "image": "image/Image collée (8).png",
+  "model": "foduucom/plant-leaf-detection-and-classification",
+  "model_source": "/home/gilles/.cache/huggingface/hub/models--foduucom--plant-leaf-detection-and-classification/snapshots/fe7b672ca4719fb7eb1b8c677c24631342a24022/best.pt",
+  "params": {
+    "conf": 0.03,
+    "iou": 0.6,
+    "max_det": 1000
+  },
+  "count": 4,
+  "detections": [
+    {
+      "index": 0,
+      "class_id": 8,
+      "class_name": "raspberry",
+      "class_name_fr": "raspberry",
+      "confidence": 0.886375367641449,
+      "bbox_xyxy": [
+        79.77804565429688,
+        5.03753137588501,
+        172.6075897216797,
+        62.644927978515625
+      ]
+    },
+    {
+      "index": 1,
+      "class_id": 8,
+      "class_name": "raspberry",
+      "class_name_fr": "raspberry",
+      "confidence": 0.8398752212524414,
+      "bbox_xyxy": [
+        161.0865478515625,
+        54.80467987060547,
+        233.95501708984375,
+        127.42171478271484
+      ]
+    },
+    {
+      "index": 2,
+      "class_id": 8,
+      "class_name": "raspberry",
+      "class_name_fr": "raspberry",
+      "confidence": 0.8144381046295166,
+      "bbox_xyxy": [
+        86.82860565185547,
+        82.79867553710938,
+        171.43515014648438,
+        195.80043029785156
+      ]
+    },
+    {
+      "index": 3,
+      "class_id": 8,
+      "class_name": "raspberry",
+      "class_name_fr": "raspberry",
+      "confidence": 0.5520069599151611,
+      "bbox_xyxy": [
+        165.154052734375,
+        9.37171459197998,
+        231.52166748046875,
+        60.80621337890625
+      ]
+    }
+  ]
+}

test_yolo/test_yolo_leaf.py

@@ -0,0 +1,169 @@
+#!/usr/bin/env python3
+from __future__ import annotations
+
+import argparse
+import json
+from pathlib import Path
+
+
+TOKEN_FR = {
+    "healthy": "saine",
+    "leaf": "feuille",
+    "plant": "plante",
+    "disease": "maladie",
+    "blight": "mildiou",
+    "rust": "rouille",
+    "spot": "tache",
+    "scab": "tavelure",
+    "mold": "moisissure",
+    "mildew": "oidium",
+    "powdery": "poudreux",
+    "bacterial": "bacterien",
+    "viral": "viral",
+    "virus": "virus",
+    "yellow": "jaune",
+    "mosaic": "mosaique",
+    "late": "tardif",
+    "early": "precoce",
+    "septoria": "septoriose",
+    "target": "cible",
+    "black": "noire",
+    "rot": "pourriture",
+}
+
+
+def translate_label_to_fr(label: str) -> str:
+    tokens = (
+        label.replace("-", " ")
+        .replace("_", " ")
+        .replace(",", " ")
+        .replace("(", " ")
+        .replace(")", " ")
+        .split()
+    )
+    if not tokens:
+        return label
+    translated = [TOKEN_FR.get(tok.lower(), tok.lower()) for tok in tokens]
+    return " ".join(translated)
+
+
+def resolve_model(model_arg: str) -> tuple[str, str]:
+    model_path = Path(model_arg)
+    if model_path.exists():
+        return str(model_path), model_arg
+
+    lower = model_arg.lower()
+    if "/" in model_arg and not lower.endswith((".pt", ".onnx", ".engine", ".torchscript")):
+        try:
+            from huggingface_hub import hf_hub_download
+        except Exception as exc:
+            raise SystemExit(
+                "Le modèle semble être un repo Hugging Face. Installe:\n"
+                "pip install huggingface_hub\n"
+                f"Détail: {exc}"
+            )
+
+        try:
+            downloaded = hf_hub_download(repo_id=model_arg, filename="best.pt")
+            return downloaded, model_arg
+        except Exception as exc:
+            raise SystemExit(
+                f"Impossible de télécharger best.pt depuis le repo Hugging Face '{model_arg}'.\n"
+                f"Détail: {exc}"
+            )
+
+    return model_arg, model_arg
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(
+        description="Test YOLO feuille/plante avec sortie JSON et image annotée."
+    )
+    parser.add_argument("--image", required=True, help="Chemin image à analyser")
+    parser.add_argument(
+        "--model",
+        default="foduucom/plant-leaf-detection-and-classification",
+        help="Chemin vers .pt/.onnx ou repo Hugging Face (best.pt)",
+    )
+    parser.add_argument("--conf", type=float, default=0.25, help="Seuil confiance")
+    parser.add_argument("--iou", type=float, default=0.45, help="Seuil IoU NMS")
+    parser.add_argument("--max-det", type=int, default=1000, help="Nombre max de détections")
+    parser.add_argument(
+        "--json-out",
+        default="test_yolo/output/detections.json",
+        help="Fichier JSON de sortie",
+    )
+    parser.add_argument(
+        "--image-out",
+        default="test_yolo/output/annotated.jpg",
+        help="Image annotée de sortie",
+    )
+    args = parser.parse_args()
+
+    image_path = Path(args.image)
+    if not image_path.exists():
+        raise SystemExit(f"Image introuvable: {image_path}")
+
+    try:
+        import cv2
+        from ultralytics import YOLO
+    except Exception as exc:
+        raise SystemExit(
+            "Dépendances manquantes. Installe dans un venv:\n"
+            "pip install ultralytics opencv-python matplotlib huggingface_hub\n"
+            f"Détail: {exc}"
+        )
+
+    model_source, model_display = resolve_model(args.model)
+    model = YOLO(model_source)
+    model.overrides["conf"] = args.conf
+    model.overrides["iou"] = args.iou
+    model.overrides["max_det"] = args.max_det
+
+    results = model.predict(str(image_path))
+    result = results[0]
+
+    boxes = result.boxes
+    names = getattr(result, "names", {})
+    detections: list[dict] = []
+    for i, box in enumerate(boxes):
+        cls_id = int(box.cls.item())
+        conf = float(box.conf.item())
+        xyxy = [float(v) for v in box.xyxy[0].tolist()]
+        detections.append(
+            {
+                "index": i,
+                "class_id": cls_id,
+                "class_name": names.get(cls_id, str(cls_id)),
+                "class_name_fr": translate_label_to_fr(names.get(cls_id, str(cls_id))),
+                "confidence": conf,
+                "bbox_xyxy": xyxy,
+            }
+        )
+
+    payload = {
+        "image": str(image_path),
+        "model": model_display,
+        "model_source": model_source,
+        "params": {"conf": args.conf, "iou": args.iou, "max_det": args.max_det},
+        "count": len(detections),
+        "detections": detections,
+    }
+
+    json_out = Path(args.json_out)
+    json_out.parent.mkdir(parents=True, exist_ok=True)
+    json_out.write_text(json.dumps(payload, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    annotated = result.plot()
+    image_out = Path(args.image_out)
+    image_out.parent.mkdir(parents=True, exist_ok=True)
+    cv2.imwrite(str(image_out), annotated)
+
+    print(f"Détections: {len(detections)}")
+    print(f"JSON: {json_out}")
+    print(f"Image annotée: {image_out}")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())