Create app.py

app.py

@@ -0,0 +1,71 @@
+import torch
+import open_clip
+import gradio as gr
+from PIL import Image
+
+# Device
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+# Load model
+model, _, preprocess = open_clip.create_model_and_transforms(
+    'hf-hub:microsoft/BiomedCLIP-PubMedBERT_256-vit_base_patch16_224'
+)
+
+tokenizer = open_clip.get_tokenizer(
+    'hf-hub:microsoft/BiomedCLIP-PubMedBERT_256-vit_base_patch16_224'
+)
+
+model = model.to(device).eval()
+
+# Labels
+labels = [
+    "CT scan of normal lung tissue",
+    "CT scan showing lung adenocarcinoma",
+    "CT scan showing large cell carcinoma",
+    "CT scan showing squamous cell carcinoma"
+]
+
+class_names = [
+    "adenocarcinoma_left.lower.lobe_T2_N0_M0_Ib",
+    "large.cell.carcinoma_left.hilum_T2_N2_M0_IIIa",
+    "normal",
+    "squamous.cell.carcinoma_left.hilum_T1_N2_M0_IIIa"]
+
+# Prediction function
+def predict(image):
+    image = preprocess(image.convert("RGB")).unsqueeze(0).to(device)
+    text = tokenizer(labels).to(device)
+
+    with torch.no_grad():
+        img_feat = model.encode_image(image)
+        txt_feat = model.encode_text(text)
+
+        img_feat /= img_feat.norm(dim=-1, keepdim=True)
+        txt_feat /= txt_feat.norm(dim=-1, keepdim=True)
+
+        similarity = img_feat @ txt_feat.T
+        probs = similarity.softmax(dim=-1)[0]
+
+    pred = torch.argmax(probs).item()
+    return class_names[pred],probs.cpu().tolist()
+
+def app(image):
+    pred, probs = predict(image)
+
+    return {
+        "Prediction": pred,
+        "Normal": float(probs[0]),
+        "adenocarcinoma_left.lower.lobe_T2_N0_M0_Ib": float(probs[1]),
+        "large.cell.carcinoma_left.hilum_T2_N2_M0_IIIa": float(probs[2]),
+        "squamous.cell.carcinoma_left.hilum_T1_N2_M0_IIIa": float(probs[3])
+    }
+
+demo = gr.Interface(
+    fn=app,
+    inputs=gr.Image(type="pil"),
+    outputs="json",
+    title="Multimodal Lung Cancer AI",
+    description="Upload a CT scan image to classify lung cancer type using a vision-language AI model."
+)
+
+demo.launch()