Upload 2 files

app.py

@@ -0,0 +1,84 @@
+import gradio as gr
+import tensorflow as tf
+import numpy as np
+from PIL import Image
+
+# 1. Load the models
+# Using try-except so the Space doesn't crash if models aren't fully uploaded yet
+try:
+    mri_model = tf.keras.models.load_model("mri_model.keras")
+    xray_model = tf.keras.models.load_model("xray_model.h5")
+except Exception as e:
+    print(f"Error loading models: {e}")
+    mri_model, xray_model = None, None
+
+# 2. Define the labels based on your training data
+mri_labels = ['Glioma', 'Meningioma', 'Pituitary tumor', 'no tumor']
+xray_labels = [
+    'Cardiomegaly', 'Emphysema', 'Effusion', 'Hernia', 'Infiltration', 
+    'Mass', 'Nodule', 'Atelectasis', 'Pneumothorax', 'Pleural_Thickening', 
+    'Pneumonia', 'Fibrosis', 'Edema', 'Consolidation'
+]
+
+# 3. Prediction Function for MRI
+def predict_mri(img):
+    if mri_model is None: return {"Error": 0.0}
+    if img is None: return {"No image": 0.0}
+    
+    # Resize to 256x256
+    img = img.resize((256, 256))
+    img_array = np.array(img)
+    
+    # Ensure image has 3 color channels (RGB)
+    if len(img_array.shape) == 2:  # If grayscale
+        img_array = np.stack((img_array,)*3, axis=-1)
+        
+    img_array = np.expand_dims(img_array, axis=0) # Add batch dimension
+    
+    # NOTE: If you trained your model with normalization (e.g. dividing by 255), uncomment the next line:
+    # img_array = img_array / 255.0
+    
+    prediction = mri_model.predict(img_array)[0]
+    confidences = {mri_labels[i]: float(prediction[i]) for i in range(len(mri_labels))}
+    return confidences
+
+# 4. Prediction Function for X-Ray
+def predict_xray(img):
+    if xray_model is None: return {"Error": 0.0}
+    if img is None: return {"No image": 0.0}
+    
+    # Resize to 128x128
+    img = img.resize((128, 128))
+    img_array = np.array(img)
+    
+    # Ensure image has 3 color channels (RGB)
+    if len(img_array.shape) == 2:  # If grayscale
+        img_array = np.stack((img_array,)*3, axis=-1)
+        
+    img_array = np.expand_dims(img_array, axis=0)
+    
+    # NOTE: If you trained your model with normalization, uncomment the next line:
+    # img_array = img_array / 255.0
+    
+    prediction = xray_model.predict(img_array)[0]
+    confidences = {xray_labels[i]: float(prediction[i]) for i in range(len(xray_labels))}
+    return confidences
+
+# 5. Build the UI / API
+with gr.Blocks() as demo:
+    gr.Markdown("# Medical Image Classification API")
+    
+    with gr.Tab("MRI Classifier"):
+        mri_input = gr.Image(type="pil", label="Upload MRI Image")
+        # num_top_classes=1 ensures we only return the highest confidence score
+        mri_output = gr.Label(num_top_classes=1, label="Result") 
+        mri_btn = gr.Button("Predict MRI")
+        mri_btn.click(fn=predict_mri, inputs=mri_input, outputs=mri_output, api_name="predict_mri")
+        
+    with gr.Tab("X-Ray Classifier"):
+        xray_input = gr.Image(type="pil", label="Upload X-Ray Image")
+        xray_output = gr.Label(num_top_classes=1, label="Result")
+        xray_btn = gr.Button("Predict X-Ray")
+        xray_btn.click(fn=predict_xray, inputs=xray_input, outputs=xray_output, api_name="predict_xray")
+
+demo.launch()

requirements.txt

@@ -0,0 +1,4 @@
+tensorflow
+numpy
+pillow
+gradio