Update app/predict.py

app/predict.py

@@ -12,41 +12,55 @@ from app.utils import clean_name, risk_from_prob, generate_leaf_report
 BASE_DIR = os.path.dirname(os.path.abspath(__file__))
 CLASSES_PATH = os.path.join(BASE_DIR, "..", "classes.json")
 
+
+
 with open(CLASSES_PATH, "r") as f:
-    CLASS_NAMES = json.load(f)
+   CLASS_NAMES = json.load(f)
+
+
 
 model = load_model()
 
 
 def predict_pil_image(pil_image: Image.Image):
-    img = pil_image.convert("RGB")
-    input_tensor = model.transform(img).unsqueeze(0)
-
-    with torch.no_grad():
-        logits = model.model(input_tensor)
-        probs = torch.softmax(logits, dim=1)[0].cpu().numpy()
-
-    pred_idx = int(np.argmax(probs))
-    pred_name = clean_name(CLASS_NAMES[pred_idx])
-    confidence = round(float(probs[pred_idx]) * 100, 2)
-
-    risk = risk_from_prob(probs[pred_idx])
-    description, treatment = generate_leaf_report(pred_name)
-
-    top5_idx = probs.argsort()[-5:][::-1]
-    top5 = [
-        {
-            "class": clean_name(CLASS_NAMES[i]),
-            "probability": float(probs[i])
-        }
-        for i in top5_idx
-    ]
-
-    return {
-        "prediction": pred_name,
-        "confidence": confidence,
-        "risk": risk,
-        "description": description,
-        "treatment": treatment,
-        "top5": top5
-    }
+   img = pil_image.convert("RGB")
+   input_tensor = model.transform(img).unsqueeze(0)
+
+
+
+   with torch.no_grad():
+       logits = model.model(input_tensor)
+       probs = torch.softmax(logits, dim=1)[0].cpu().numpy()
+
+
+
+   pred_idx = int(np.argmax(probs))
+   pred_name = clean_name(CLASS_NAMES[pred_idx])
+   confidence = round(float(probs[pred_idx]) * 100, 2)
+
+
+
+   risk = risk_from_prob(probs[pred_idx])
+   description, treatment = generate_leaf_report(pred_name)
+
+
+
+   top5_idx = probs.argsort()[-5:][::-1]
+   top5 = [
+       {
+           "class": clean_name(CLASS_NAMES[i]),
+           "probability": float(probs[i])
+       }
+       for i in top5_idx
+   ]
+
+
+
+   return {
+       "prediction": pred_name,
+       "confidence": confidence,
+       "risk": risk,
+       "description": description,
+       "treatment": treatment,
+       "top5": top5
+   }