Update app.py

app.py

@@ -4,6 +4,7 @@ import inspect
 import sys
 import os
 import types
+import shutil
 from unittest.mock import MagicMock
 
 import numpy as np
@@ -23,7 +24,7 @@ from fastapi.responses import JSONResponse
 import uvicorn
 from huggingface_hub import hf_hub_download
 
-# --- Compatibility Patches ---
+# --- Compatibility Patches for Numpy and Inspect ---
 if not hasattr(inspect, "getargspec"):
     inspect.getargspec = inspect.getfullargspec
 
@@ -32,7 +33,7 @@ for attr, typ in [("int", int), ("float", float), ("complex", complex),
     if not hasattr(np, attr):
         setattr(np, attr, typ)
 
-# --- Pyrender / OpenGL Mock (Headless Fix) ---
+# --- Pyrender / OpenGL Mock (Headless Environment Fix) ---
 pyrender_mock = types.ModuleType("pyrender")
 for _attr in ["Scene", "Mesh", "Node", "PerspectiveCamera", "DirectionalLight",
               "PointLight", "SpotLight", "OffscreenRenderer", "RenderFlags",
@@ -50,14 +51,35 @@ os.environ["PYOPENGL_PLATFORM"] = "osmesa"
 # --- Hugging Face Model Integration ---
 REPO_ID = "SondosM/api_GP"
 
-def get_hf_file(filename):
-    return hf_hub_download(repo_id=REPO_ID, filename=filename)
+def get_hf_file(filename, is_mano=False):
+    print(f"Downloading {filename} from {REPO_ID}...")
+    temp_path = hf_hub_download(repo_id=REPO_ID, filename=filename)
+    
+    if is_mano:
+        # Create local folder structure expected by WiLoR
+        os.makedirs("./mano_data", exist_ok=True)
+        target_path = os.path.join("./mano_data", os.path.basename(filename))
+        if not os.path.exists(target_path):
+            shutil.copy(temp_path, target_path)
+            print(f"Copied {filename} to {target_path}")
+        return target_path
+    
+    return temp_path
+
+# --- Map paths according to your Repo list ---
+print("Initializing model file paths...")
+
+# MANO Files
+get_hf_file("mano_data/mano_data/mano_mean_params.npz", is_mano=True)
+get_hf_file("mano_data/mano_data/MANO_LEFT.pkl", is_mano=True)
+get_hf_file("mano_data/mano_data/MANO_RIGHT.pkl", is_mano=True)
 
-# Resolve paths from Hugging Face Repo
 WILOR_REPO_PATH   = "./WiLoR"
+# Model weights
 WILOR_CKPT        = get_hf_file("pretrained_models/pretrained_models/wilor_final.ckpt")
 WILOR_CFG         = get_hf_file("pretrained_models/pretrained_models/model_config.yaml")
 DETECTOR_PATH     = get_hf_file("pretrained_models/pretrained_models/detector.pt")
+# Classifier
 CLASSIFIER_PATH   = get_hf_file("classifier.pkl")
 
 DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
@@ -82,12 +104,12 @@ def load_models():
     wilor_model.to(DEVICE)
     wilor_model.eval()
 
-    print(f"Loading YOLO from: {DETECTOR_PATH}")
+    print(f"Loading YOLO detector...")
     yolo_detector = YOLO(DETECTOR_PATH)
 
-    print("Loading Classifier...")
+    print("Loading RandomForest classifier...")
     classifier = joblib.load(CLASSIFIER_PATH)
-    print("Models loaded!")
+    print("✅ All models loaded successfully!")
 
 @asynccontextmanager
 async def lifespan(app: FastAPI):
@@ -141,7 +163,7 @@ def read_image_from_upload(file_bytes: bytes) -> np.ndarray:
     arr = np.frombuffer(file_bytes, np.uint8)
     img = cv2.imdecode(arr, cv2.IMREAD_COLOR)
     if img is None:
-        raise HTTPException(status_code=400, detail="Invalid image.")
+        raise HTTPException(status_code=400, detail="Invalid image format.")
     return img
 
 @app.get("/")
@@ -168,7 +190,7 @@ async def predict(file: UploadFile = File(...)):
     crop = img_rgb[y1:y2, x1:x2]
 
     if crop.size == 0:
-        raise HTTPException(status_code=422, detail="Empty crop.")
+        raise HTTPException(status_code=422, detail="Empty hand crop.")
 
     features = extract_features(crop)
     if features is None:
@@ -184,7 +206,7 @@ async def predict(file: UploadFile = File(...)):
     proba = classifier.predict_proba(feat_df)[0]
 
     return JSONResponse({
-        "prediction": prediction,
+        "prediction": str(prediction),
         "confidence": round(float(proba.max()), 4),
         "hand_side": hand_side,
         "bbox": [int(x1), int(y1), int(x2), int(y2)],
@@ -224,7 +246,7 @@ async def predict_with_skeleton(file: UploadFile = File(...)):
     crop_b64 = base64.b64encode(buf).decode("utf-8")
 
     return JSONResponse({
-        "prediction": prediction,
+        "prediction": str(prediction),
         "confidence": round(float(proba.max()), 4),
         "hand_side": hand_side,
         "bbox": [int(x1), int(y1), int(x2), int(y2)],
@@ -233,5 +255,4 @@ async def predict_with_skeleton(file: UploadFile = File(...)):
     })
 
 if __name__ == "__main__":
-    # Hugging Face Spaces port is 7860
     uvicorn.run("app:app", host="0.0.0.0", port=7860, reload=False)