all problems are now fixed. \detect should return predictions

README.md

@@ -1,14 +1,19 @@
+# YOLO FastAPI Deploy 🚀
+
+This Space deploys a YOLO object detection model with **FastAPI** on Hugging Face Spaces.
+
+## 🔗 Live URL
+
+👉 [https://ujoy-yolo-fast-api-deploy.hf.space](https://ujoy-yolo-fast-api-deploy.hf.space)
+
+- Root health: `/`
+- API docs (Swagger UI): `/docs`
+- Detect endpoint: `/detect`
+
 ---
-title: YOLO FastAPI Deploy
-emoji: 🤖
-colorFrom: blue
-colorTo: green
-sdk: docker
-app_file: main.py
-pinned: false
----
 
-# YOLO FastAPI Deployment 🚀
+## 📌 Usage
+
+### 1. Upload an image for detection
 
-This Space deploys a YOLO model with FastAPI.  
-Use `/docs` to test endpoints, or send requests from Postman.
+**Endpoint**:

main.py

@@ -1,32 +1,33 @@
 from fastapi import FastAPI, UploadFile, File
-from fastapi.responses import JSONResponse, FileResponse
+from fastapi.responses import JSONResponse
 from ultralytics import YOLO
 import shutil
 import uuid
 import os
-import base64
 
 app = FastAPI()
 
 # Load YOLO model
 model = YOLO("yolo11s.pt")
 
-# Use /tmp for writable storage in Hugging Face Spaces
+# Writable directory on Hugging Face
 UPLOAD_DIR = "/tmp/uploads"
 os.makedirs(UPLOAD_DIR, exist_ok=True)
 
 
 @app.get("/")
 def home():
-    return {"message": "YOLO FastAPI is running! 🚀 Go to /docs for the API."}
+    return {"status": "running", "docs": "/docs"}
 
 
 @app.post("/detect")
-async def detect(file: UploadFile = File(...), return_type: str = "json"):
+async def detect(file: UploadFile = File(...)):
     """
     Run YOLO detection.
-    - return_type=json → return base64 encoded image inside JSON
-    - return_type=file → return annotated image as downloadable file
+    Returns:
+    - bounding boxes
+    - class labels
+    - confidence scores
     """
     # Save uploaded file
     file_id = str(uuid.uuid4())
@@ -37,19 +38,13 @@ async def detect(file: UploadFile = File(...), return_type: str = "json"):
     # Run YOLO detection
     results = model(file_path)
 
-    # Save annotated result image
-    result_path = f"{file_path}_result.jpg"
-    results[0].save(filename=result_path)
-
-    if return_type == "file":
-        # Return file directly (downloadable)
-        return FileResponse(result_path, media_type="image/jpeg", filename="result.jpg")
-    else:
-        # Encode image as base64 string
-        with open(result_path, "rb") as img_file:
-            encoded = base64.b64encode(img_file.read()).decode("utf-8")
-
-        return JSONResponse(content={
-            "message": "Detection complete",
-            "image_base64": encoded
+    # Extract predictions
+    predictions = []
+    for box in results[0].boxes:
+        predictions.append({
+            "class": results[0].names[int(box.cls)],
+            "confidence": float(box.conf),
+            "bbox": [float(x) for x in box.xyxy[0].tolist()]  # [x1, y1, x2, y2]
         })
+
+    return JSONResponse(content={"predictions": predictions})