Update app.py

app.py

@@ -1,34 +1,56 @@
-import gradio as gr
-import pandas as pd
 import numpy as np
-from tensorflow.keras.models import load_model
+import gradio as gr
 from PIL import Image
-from sklearn.preprocessing import LabelEncoder
-
-# Load the trained model
-model = load_model("best_dog_breed_model.keras")  # ✅ match the actual file name
-
-# Load and process labels
-df = pd.read_csv("labels.csv")
-le = LabelEncoder()
-le.fit(df['breed'])
-breed_list = list(le.classes_)
-
-# Define prediction function
-def predict(image):
-    image = image.resize((224, 224))
-    img_array = np.expand_dims(np.array(image) / 255.0, axis=0)
-    preds = model.predict(img_array)[0]
-    top5_idx = preds.argsort()[-5:][::-1]
-    return {breed_list[i]: float(preds[i]) for i in top5_idx}
-
-# Create Gradio interface
+import tensorflow as tf
+
+# Load your model here
+model = tf.keras.models.load_model('your_model.h5')
+
+# List of classes your model predicts
+class_names = ['beagle', 'bulldog', 'golden_retriever', 'toy', 'human']  # add all your classes here
+
+# Classes to flag as unknown
+unknown_classes = {'toy', 'human'}
+
+# Image size expected by your model
+IMAGE_SIZE = (224, 224)
+
+def preprocess_image(img):
+    img = img.convert("RGB").resize(IMAGE_SIZE)
+    img = np.array(img) / 255.0
+    img = np.expand_dims(img, axis=0)
+    return img
+
+def predict(img):
+    img_processed = preprocess_image(img)
+    preds = model.predict(img_processed)[0]  # assuming batch size 1
+    top5_indices = preds.argsort()[-5:][::-1]
+    top5_preds = [(class_names[i], preds[i]) for i in top5_indices]
+
+    # Check unknown classes in top 5
+    if any(cls in unknown_classes for cls, conf in top5_preds):
+        return "Unknown class detected, unable to classify reliably."
+
+    # Final prediction with threshold (e.g. 90%)
+    final_idx = np.argmax(preds)
+    confidence = preds[final_idx]
+    if confidence < 0.9:
+        return "Confidence too low to predict reliably."
+
+    final_breed = class_names[final_idx]
+    result_str = f"Final Prediction: {final_breed} (Confidence: {confidence:.2f})\n\nTop 5 predictions:\n"
+    for cls, conf in top5_preds:
+        result_str += f"{cls}: {conf:.2f}\n"
+
+    return result_str
+
 iface = gr.Interface(
     fn=predict,
     inputs=gr.Image(type="pil"),
-    outputs=gr.Label(num_top_classes=5),
+    outputs=gr.Textbox(),
     title="Dog Breed Classifier",
-    description="Upload a dog image to see the top 5 predicted breeds."
+    description="Upload an image and see top 5 predictions and final breed with threshold and unknown detection."
 )
 
-iface.launch()
+if __name__ == "__main__":
+    iface.launch()