added bounding box features

Dockerfile

@@ -13,6 +13,8 @@ COPY --chown=user ./requirements.txt requirements.txt
 RUN pip install --no-cache-dir --upgrade -r requirements.txt
 RUN pip install torch torchvision --index-url https://download.pytorch.org/whl/cpu
 RUN pip install torchxrayvision
+RUN pip install grad-cam
+RUN pip install opencv-python
 
 COPY --chown=user . /app
 CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]

app.py

@@ -1,11 +1,64 @@
 from fastapi import FastAPI,Query,HTTPException
 import torchxrayvision as xrv
 import skimage, torch, torchvision
-import requests
+import cv2
+import numpy as np
+from pytorch_grad_cam import GradCAM
+from pytorch_grad_cam.utils.model_targets import ClassifierOutputTarget
 
 
+from fastapi.middleware.cors import CORSMiddleware
 app = FastAPI()
+# Add the frontend origin here
+origins = [
+    "http://localhost:8080",  # Your frontend running on port 8080
+    "http://127.0.0.1:8080"
+]
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=origins,  # OR ["*"] only during dev
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+
 model = xrv.models.DenseNet(weights="densenet121-res224-all")
+
+
+
+
+
+def show_anomaly_bounding_box(img_tensor, model, class_index=None):
+    target_layer = model.features[-1]
+
+    cam = GradCAM(model=model, target_layers=[target_layer])
+
+    with torch.no_grad():
+        outputs = model(img_tensor[None, ...])
+        pred_index = class_index if class_index is not None else torch.argmax(outputs[0]).item()
+
+    grayscale_cam = cam(input_tensor=img_tensor[None, ...],
+                        targets=[ClassifierOutputTarget(pred_index)])
+    grayscale_cam = grayscale_cam[0, :]
+
+    input_img = img_tensor.numpy()[0]
+    input_img_norm = (input_img - input_img.min()) / (input_img.max() - input_img.min())
+    input_img_rgb = cv2.cvtColor((input_img_norm * 255).astype(np.uint8), cv2.COLOR_GRAY2RGB)
+
+    cam_resized = cv2.resize(grayscale_cam, (224, 224))
+    cam_uint8 = (cam_resized * 255).astype(np.uint8)
+    _, thresh = cv2.threshold(cam_uint8, 100, 255, cv2.THRESH_BINARY)
+    contours, _ = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
+    bounding_box = ()
+    for cnt in contours:
+        x, y, w, h = cv2.boundingRect(cnt)
+        bounding_box = ((x,y),(x+w,y+h))
+        # cv2.rectangle(input_img_rgb, (x, y), (x + w, y + h), (0, 255, 0), 2)
+
+    return bounding_box
+
 @app.get("/")
 def greet_json():
     return {"Hello": "World!"}
@@ -28,8 +81,9 @@ def predict(image_url:str = Query(..., description="URL to a chest X-ray image")
         for k,v in prediction.items():
             pred_output.update({k:round(v,2)})
 
+        get_bounding_box = show_anomaly_bounding_box(img,model=model)
 
-        return {"prediction_result":pred_output,"bounding_box":{pred_label:((139,61),(224,132))}}
+        return {"prediction_result":pred_output,"bounding_box":{pred_label:get_bounding_box}}
     except Exception as e:
         print(e)
         raise HTTPException(status_code=400, detail=f"Failed to fetch/process image: {str(e)}")