Fix added

SkinCancerDiagnosis.py

@@ -139,6 +139,16 @@ class SkinDiseaseClassifier:
             transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
         ])
 
+        self.multilabel_class_names = [
+            "Vesicle", "Papule", "Macule", "Plaque", "Abscess", "Pustule", "Bulla", "Patch",
+            "Nodule", "Ulcer", "Crust", "Erosion", "Excoriation", "Atrophy", "Exudate", "Purpura/Petechiae",
+            "Fissure", "Induration", "Xerosis", "Telangiectasia", "Scale", "Scar", "Friable", "Sclerosis",
+            "Pedunculated", "Exophytic/Fungating", "Warty/Papillomatous", "Dome-shaped", "Flat topped",
+            "Brown(Hyperpigmentation)", "Translucent", "White(Hypopigmentation)", "Purple", "Yellow",
+            "Black", "Erythema", "Comedo", "Lichenification", "Blue", "Umbilicated", "Poikiloderma",
+            "Salmon", "Wheal", "Acuminate", "Burrow", "Gray", "Pigmented", "Cyst"
+        ]
+
     def load_models(self):
         """Load all required models"""
         # Load binary models
@@ -179,6 +189,18 @@ class SkinDiseaseClassifier:
         self.meta_model.to(self.device)
         self.meta_model.eval()
 
+        skincon_path = hf_hub_download(
+            repo_id="KeerthiVM/SkinCancerDiagnosis",
+            filename="skincon.pth"
+        )
+        self.skincon_model = EvoViTModel(img_size=224, patch_size=16, in_channels=3, embed_dim=768, num_classes=48,
+                                         hidden_dim=512)
+        state_dict = torch.load(skincon_path, map_location=device)
+        self.skincon_model.load_state_dict(state_dict, strict=False)
+
+        self.skincon_model.eval()
+
+
     def extract_image_features(self, image_tensor):
         """Extract features using ResNet"""
         with torch.no_grad():
@@ -197,12 +219,10 @@ class SkinDiseaseClassifier:
     def predict(self, image, top_k=3):
         """Make prediction for a single image"""
         if self.base_models is None or self.meta_model is None:
-            # self.load_models()
             raise RuntimeError("Models not loaded - call load_models() first")
 
         # Load and preprocess image
         try:
-            # image = Image.open(image_path).convert('RGB')
             image = image.convert('RGB')
         except:
             raise ValueError("Could not load image from path")
@@ -257,6 +277,27 @@ class SkinDiseaseClassifier:
             "all_probabilities": {name: float(prob) for name, prob in zip(self.class_names, probabilities)}
         }
 
+    def predict_skincon(self, image, top_k=3):
+        """Make prediction for a single image"""
+        if self.base_models is None or self.skincon_model is None:
+            raise RuntimeError("Models not loaded - call load_models() first")
+        self.skincon_model.eval()
+        try:
+            image = image.convert('RGB')
+        except:
+            raise ValueError("Could not load image from path")
+
+        image_tensor = self.transform(image).unsqueeze(0).to(self.device)
+        with torch.no_grad():
+            output_multi = self.skincon_model(image_tensor)
+            probs_multi = torch.sigmoid(output_multi).squeeze().numpy()
+            # print(f"Probabilities : {probs_multi}")
+            threshold = 0.5
+            predicted_labels_multi = [self.multilabel_class_names[i] for i, p in enumerate(probs_multi) if p > threshold]
+        print("Predicted labels multi : ",predicted_labels_multi)
+        return predicted_labels_multi
+
+
 def initialize_classifier():
     print("⚙️ Initializing skin disease classifier...")
     classifier = SkinDiseaseClassifier()

app.py

@@ -102,7 +102,8 @@ if "current_image" not in st.session_state:
 def run_inference(image):
     result = classifier.predict(image, top_k=1)
     predicted_label = result["top_predictions"][0][0]
-    return predicted_label
+    predicted_label_multi = classifier.predict_skincon(image, top_k=1)
+    return predicted_label, predicted_label_multi
 
 
 # === PDF Export ===
@@ -136,11 +137,12 @@ if uploaded_file is not None and uploaded_file != st.session_state.current_image
     image = Image.open(uploaded_file).convert("RGB")
     st.image(image, caption="Uploaded image", use_column_width=True)
     with st.spinner("Analyzing the image..."):
-        predicted_label = run_inference(image)
+        predicted_label, predicted_label_multi = run_inference(image)
 
+    st.markdown(f"🧾 **Skin Issues**: {', '.join(predicted_label_multi)}")
     st.markdown(f" Most Likely Diagnosis : {predicted_label}")
 
-    initial_query = f"What are my treatment options for {predicted_label}?"
+    initial_query = f"What are my treatment options for {predicted_label} & {predicted_label_multi}?"
     st.session_state.messages.append({"role": "user", "content": initial_query})
     with st.spinner("Retrieving medical information..."):
         response = get_reranked_response(initial_query, st.session_state.app_models['llm'], st.session_state.app_models['rag_components'])