Create app.py

app.py

@@ -0,0 +1,97 @@
+import base64
+import cv2
+import numpy as np
+from fastapi import FastAPI
+from pydantic import BaseModel
+from collections import defaultdict
+import uvicorn
+import os
+
+app = FastAPI()
+
+class InputData(BaseModel):
+    image_base64: str
+
+def save_base64_image_cv(base64_str, output_path="final.png"):
+    image_data = base64.b64decode(base64_str)
+    nparr = np.frombuffer(image_data, np.uint8)
+    img = cv2.imdecode(nparr, cv2.IMREAD_UNCHANGED)
+
+    # jika RGBA → tempelkan ke background putih
+    if img is not None and img.shape[2] == 4:
+        alpha = img[:, :, 3] / 255.0
+        rgb = img[:, :, :3]
+        white_bg = np.ones_like(rgb, dtype=np.uint8) * 255
+        img = (rgb * alpha[:, :, None] + white_bg * (1 - alpha[:, :, None])).astype(np.uint8)
+
+    cv2.imwrite(output_path, img)
+    return output_path
+
+def extract_icon_positions(image_path="final.png"):
+    img = cv2.imread(image_path)
+    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+
+    _, thresh = cv2.threshold(gray, 200, 255, cv2.THRESH_BINARY_INV)
+    contours, _ = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
+
+    icon_features = []
+    positions = []
+
+    for cnt in contours:
+        x, y, w, h = cv2.boundingRect(cnt)
+        if w > 10 and h > 10:
+            roi = thresh[y:y+h, x:x+w]
+            roi_resized = cv2.resize(roi, (50, 50))
+            icon_features.append(roi_resized)
+            positions.append((x, y))
+
+    return icon_features, positions
+
+def img_hash(img):
+    resized = cv2.resize(img, (8, 8))
+    mean = resized.mean()
+    return (resized > mean).astype(np.uint8).flatten()
+
+def find_rarest_icon_position(icon_features, positions):
+    hashes = [img_hash(icon) for icon in icon_features]
+    groups = defaultdict(list)
+
+    for i, h in enumerate(hashes):
+        found = False
+        for label, group in groups.items():
+            dist = np.sum(h != hashes[group[0]])
+            if dist < 3:
+                group.append(i)
+                found = True
+                break
+        if not found:
+            groups[len(groups)] = [i]
+
+    rarest_group = min(groups.values(), key=len)
+    idx = rarest_group[0]
+
+    return positions[idx][0], positions[idx][1]
+
+@app.post("/predict")
+async def predict(data: InputData):
+    try:
+        save_base64_image_cv(data.image_base64, "final.png")
+
+        icon_features, positions = extract_icon_positions("final.png")
+
+        if not positions:
+            return {"error": "No icons detected"}
+
+        x, y = find_rarest_icon_position(icon_features, positions)
+
+        # hapus file temp
+        if os.path.exists("final.png"):
+            os.remove("final.png")
+
+        return {"x": x, "y": y}
+
+    except Exception as e:
+        return {"error": str(e)}
+
+if __name__ == "__main__":
+    uvicorn.run(app, host="0.0.0.0", port=7860)