Create app.py

app.py

@@ -0,0 +1,93 @@
+from fastapi import FastAPI, UploadFile, File
+from fastapi.responses import JSONResponse
+import numpy as np
+import joblib
+import cv2
+from PIL import Image
+import mediapipe as mp
+import os
+import io
+
+app = FastAPI()
+
+# ---------------- LOAD MODEL ----------------
+MODEL_DIR = "model2"
+model = joblib.load(os.path.join(MODEL_DIR, "emotion_model.joblib"))
+label_encoder = joblib.load(os.path.join(MODEL_DIR, "label_encoder.joblib"))
+
+face_mesh = mp.solutions.face_mesh.FaceMesh(
+    static_image_mode=True,
+    max_num_faces=1,
+    min_detection_confidence=0.5
+)
+
+# Same feature order you used during training
+FEATURE_ORDER = [
+    # add your full list here exactly as in training
+]
+
+
+# ---------------- FEATURE EXTRACTION ----------------
+def compute_basic_features(landmarks, w, h):
+    # paste the exact function from your project here
+    raise NotImplementedError("Add your compute_basic_features() here")
+
+
+def extract_features(image):
+    img_rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+    results = face_mesh.process(img_rgb)
+
+    if not results.multi_face_landmarks:
+        return None, ["No face detected"]
+
+    landmarks = np.array([(lm.x, lm.y, lm.z) for lm in results.multi_face_landmarks[0].landmark])
+    h, w, _ = image.shape
+
+    features = compute_basic_features(landmarks, w, h)
+
+    ordered_features = [features.get(f, 0) for f in FEATURE_ORDER]
+    return np.array(ordered_features).reshape(1, -1), None
+
+
+def predict_emotion(image):
+    X, error = extract_features(image)
+    if error:
+        return {"error": error}
+
+    pred = model.predict(X)[0]
+    confidence = np.max(model.predict_proba(X))
+
+    emotion_label = label_encoder.inverse_transform([pred])[0]
+
+    return {
+        "emotion": emotion_label,
+        "confidence": float(confidence),
+    }
+
+
+# ---------------- API ENDPOINT ----------------
+@app.post("/predict")
+async def predict(image: UploadFile = File(...)):
+    try:
+        img_bytes = await image.read()
+        img = Image.open(io.BytesIO(img_bytes))
+        img = np.array(img)
+
+        # Normalize the format
+        if img.ndim == 2:
+            img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
+        elif img.shape[2] == 4:
+            img = cv2.cvtColor(img, cv2.COLOR_RGBA2BGR)
+        else:
+            img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
+
+        output = predict_emotion(img)
+        return JSONResponse(output)
+
+    except Exception as e:
+        return JSONResponse({"error": str(e)}, status_code=500)
+
+
+@app.get("/")
+def root():
+    return {"status": "API is running", "endpoint": "/predict"}