app.py fixed

Dockerfile

@@ -1,18 +1,14 @@
-FROM python:3.10-slim
-
-WORKDIR /app
+FROM python:3.10
 
 RUN apt-get update && apt-get install -y \
+    libgl1 \
+    libglib2.0-0 \
     git \
     && rm -rf /var/lib/apt/lists/*
 
-COPY requirements.txt .
+WORKDIR /app
+COPY . .
 
-RUN pip install --no-cache-dir --upgrade pip
 RUN pip install --no-cache-dir -r requirements.txt
 
-COPY app.py .
-
-EXPOSE 7860
-
-CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]
+CMD ["uvicorn","app:app","--host","0.0.0.0","--port","7860"]

app.py

@@ -1,96 +1,61 @@
-import io
+from fastapi import FastAPI, UploadFile
+from PIL import Image
 import torch
 import numpy as np
-from PIL import Image
-from fastapi import FastAPI, UploadFile, File, HTTPException
-from fastapi.responses import JSONResponse
-from contextlib import asynccontextmanager
-from transformers import AutoImageProcessor, AutoModelForDepthEstimation
+import cv2
+from torchvision import transforms
+from depth_anything_v2.dpt import DepthAnythingV2
 
-MODEL_ID = "depth-anything/Depth-Anything-V2-Metric-Indoor-Base-hf"
+app = FastAPI()
 
-model = None
-processor = None
-device = None
+# -------- LOAD MODEL --------
+device = "cuda" if torch.cuda.is_available() else "cpu"
 
+model = DepthAnythingV2.from_pretrained(
+    "LiheYoung/depth-anything-v2-base"
+).to(device)
 
-@asynccontextmanager
-async def lifespan(app: FastAPI):
-    global model, processor, device
+model.eval()
 
-    print("Loading Metric Depth Anything...")
-    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+# -------- TRANSFORM --------
+transform = transforms.Compose([
+    transforms.Resize((518, 518)),
+    transforms.ToTensor(),
+    transforms.Normalize(
+        mean=[0.485, 0.456, 0.406],
+        std=[0.229, 0.224, 0.225]
+    )
+])
 
-    processor = AutoImageProcessor.from_pretrained(MODEL_ID)
-    model = AutoModelForDepthEstimation.from_pretrained(MODEL_ID).to(device)
-    model.eval()
+@app.get("/")
+def root():
+    return {"message": "Depth Anything V2 API running"}
 
-    print(f"Model loaded on {device}")
-    yield
+# -------- DEPTH ENDPOINT --------
+@app.post("/depth")
+async def depth(file: UploadFile):
 
+    img = Image.open(file.file).convert("RGB")
 
-app = FastAPI(
-    title="Metric Depth Anything API",
-    version="1.0.0",
-    lifespan=lifespan,
-)
+    original_w, original_h = img.size
 
+    x = transform(img).unsqueeze(0).to(device)
 
-@app.get("/")
-def root():
-    return {"message": "Metric Depth API running"}
-
-
-@app.post("/predict")
-async def predict(file: UploadFile = File(...)):
-    try:
-        contents = await file.read()
-        image = Image.open(io.BytesIO(contents)).convert("RGB")
-    except Exception as e:
-        raise HTTPException(status_code=400, detail=f"Invalid image: {e}")
-
-    try:
-        inputs = processor(images=image, return_tensors="pt").to(device)
-
-        with torch.no_grad():
-            outputs = model(**inputs)
-
-        post = processor.post_process_depth_estimation(
-            outputs,
-            target_sizes=[(image.height, image.width)],
-        )
-
-        depth = post[0]["predicted_depth"].cpu().numpy()  # in meters
-
-        h, w = depth.shape
-
-        # pixel le plus proche (plus petite distance en mètres)
-        closest_idx = np.unravel_index(np.argmin(depth), depth.shape)
-        closest_y, closest_x = int(closest_idx[0]), int(closest_idx[1])
-        closest_distance = float(depth[closest_y, closest_x])
-
-        # pixel central
-        cy, cx = h // 2, w // 2
-        center_distance = float(depth[cy, cx])
-
-        return JSONResponse({
-            "image_size": {"width": w, "height": h},
-            "closest_pixel": {
-                "x": closest_x,
-                "y": closest_y,
-                "distance_meters": closest_distance,
-            },
-            "center_pixel": {
-                "x": cx,
-                "y": cy,
-                "distance_meters": center_distance,
-            },
-        })
-
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Inference error: {e}")
-
-
-if __name__ == "__main__":
-    import uvicorn
-    uvicorn.run("app:app", host="0.0.0.0", port=7860)
+    with torch.no_grad():
+        depth = model(x)
+
+    depth = depth.squeeze().cpu().numpy()
+
+    # resize depth back to original resolution
+    depth = cv2.resize(depth, (original_w, original_h))
+
+    closest_distance = float(np.min(depth))
+
+    h, w = depth.shape
+    center_distance = float(depth[h//2, w//2])
+
+    return {
+        "closest_distance": closest_distance,
+        "center_distance": center_distance,
+        "resolution": [w, h]
+    }

requirements.txt

@@ -1,8 +1,10 @@
 fastapi
 uvicorn
-torch
-transformers
-Pillow
+pillow
 numpy
-accelerate
-python-multipart
+opencv-python-headless
+torch
+torchvision
+timm
+huggingface_hub
+git+https://github.com/DepthAnything/Depth-Anything-V2.git