Update app.py

app.py

@@ -7,6 +7,7 @@ import torch
 import torchxrayvision as xrv
 import re
 import os
+from skimage.transform import resize as sk_resize
 
 # -----------------------------
 # Imaging Agent (Chest X-ray, proxy for lung cancer risk)
@@ -31,16 +32,16 @@ def imaging_agent(image_path: str):
         # TorchXRayVision normalization
         arr = xrv.datasets.normalize(arr, 4096)
 
-        # Wrap in dict for TorchXRayVision
-        sample = {"img": arr, "lab": np.array([]), "idx": np.array([])}
+        # --- Manual preprocessing instead of dict transforms ---
+        h, w = arr.shape
+        min_dim = min(h, w)
+        startx = w // 2 - (min_dim // 2)
+        starty = h // 2 - (min_dim // 2)
+        arr = arr[starty:starty+min_dim, startx:startx+min_dim]  # center crop
 
-        # Apply transforms step by step
-        sample = xrv.datasets.XRayCenterCrop()(sample)
-        sample = xrv.datasets.XRayResizer(224)(sample)
+        arr = sk_resize(arr, (224, 224), preserve_range=True)   # resize to 224x224
 
-        arr = sample["img"]  # Unpack back to array
-
-        # Convert to torch tensor
+        # Convert to tensor
         arr = torch.from_numpy(arr).unsqueeze(0).unsqueeze(0).to(DEVICE)
 
         # Inference
@@ -48,7 +49,7 @@ def imaging_agent(image_path: str):
             preds = MODEL(arr)[0]
             probs = torch.sigmoid(preds).cpu().numpy().tolist()
 
-        # Focus only on cancer-relevant lung pathologies
+        # Focus on lung pathologies relevant to cancer
         focus_labels = ["Lung Opacity", "Mass", "Nodule"]
         focus = [(l, probs[PATHOLOGIES.index(l)]) for l in focus_labels if l in PATHOLOGIES]
 
@@ -162,3 +163,4 @@ demo.launch()
 
 
 
+