Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import tensorflow as tf
|
3 |
+
from PIL import Image
|
4 |
+
import numpy as np
|
5 |
+
from tensorflow.keras.preprocessing.image import ImageDataGenerator
|
6 |
+
|
7 |
+
# Load the car brand classifier model
|
8 |
+
model_path = "car_brand_classifier_finetuned.keras"
|
9 |
+
model = tf.keras.models.load_model(model_path)
|
10 |
+
|
11 |
+
labels = ['Hyundai', 'Lexus', 'Mazda', 'Mercedes', 'Opel', 'Skoda', 'Toyota', 'Volkswagen']
|
12 |
+
|
13 |
+
# Define function for car brand classification with data augmentation
|
14 |
+
def preprocess_image(image):
|
15 |
+
image = Image.fromarray(image.astype('uint8'), 'RGB')
|
16 |
+
image = image.resize((224, 224))
|
17 |
+
image = np.array(image)
|
18 |
+
image = image / 255.0 # Normalize pixel values
|
19 |
+
return image
|
20 |
+
|
21 |
+
|
22 |
+
# Prediction function
|
23 |
+
def predict_car_brand(image):
|
24 |
+
image = preprocess_image(image)
|
25 |
+
prediction = model.predict(np.expand_dims(image, axis=0))
|
26 |
+
predicted_class = labels[np.argmax(prediction)]
|
27 |
+
confidence = np.round(np.max(prediction) * 100, 2)
|
28 |
+
result = f"Label: {predicted_class}, Confidence: {confidence}%"
|
29 |
+
return result
|
30 |
+
|
31 |
+
# Create Gradio interface
|
32 |
+
input_image = gr.Image()
|
33 |
+
output_text = gr.Textbox(label="Car Brand")
|
34 |
+
interface = gr.Interface(fn=predict_car_brand,
|
35 |
+
inputs=input_image,
|
36 |
+
outputs=output_text,
|
37 |
+
description="A car brand classifier using transfer learning and fine-tuning with EfficientNetB0.",
|
38 |
+
theme="default")
|
39 |
+
|
40 |
+
if __name__ == "__main__":
|
41 |
+
interface.launch()
|