File size: 1,124 Bytes
94506f9 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 |
import gradio as gr
import numpy as np
from tensorflow.keras.models import load_model
from tensorflow.keras.preprocessing import image
import tensorflow as tf
# Load the trained model
model = load_model("VGG.h5")
# Define class names (order from your dataset's subfolders)
class_names = ['cat', 'dog', 'wild'] # Change if your folder names differ
IMG_SIZE = 224
def predict(img):
# Preprocess the image
img = img.resize((IMG_SIZE, IMG_SIZE))
img_array = image.img_to_array(img)
img_array = img_array / 255.0 # Rescale
img_array = np.expand_dims(img_array, axis=0)
# Predict
preds = model.predict(img_array)[0]
result = {class_names[i]: float(preds[i]) for i in range(len(class_names))}
return result
# Build Gradio Interface
demo = gr.Interface(
fn=predict,
inputs=gr.Image(type="pil"),
outputs=gr.Label(num_top_classes=3),
title="Animal Face Classifier (VGG16)",
description="Upload an image of an animal face (cat, dog, or wild) and get the predicted class probabilities."
)
if __name__ == "__main__":
demo.launch()
|