Sjleeiv's picture
Update app.py
9a78e3c
Raw History Blame Contribute Delete
3.84 kB
import gradio as gr
import tensorflow as tf
from PIL import Image
import numpy as np
from tensorflow.keras import layers, models
from tensorflow.keras.applications.resnet import preprocess_input
labels_list = ['rime', 'fogsmog', 'dew', 'sandstorm', 'glaze']
def identity_block(x, filters):
x_shortcut = x
x = layers.Conv2D(filters[0], (1, 1), strides=(1, 1), padding='valid')(x)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.Conv2D(filters[1], (3, 3), strides=(1, 1), padding='same')(x)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.Conv2D(filters[2], (1, 1), strides=(1, 1), padding='valid')(x)
x = layers.BatchNormalization()(x)
x = layers.Add()([x, x_shortcut])
x = layers.Activation('relu')(x)
return x
def conv_block(x, filters, stride):
x_shortcut = x
x = layers.Conv2D(filters[0], (1, 1), strides=(stride, stride), padding='valid')(x)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.Conv2D(filters[1], (3, 3), strides=(1, 1), padding='same')(x)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.Conv2D(filters[2], (1, 1), strides=(1, 1), padding='valid')(x)
x = layers.BatchNormalization()(x)
x_shortcut = layers.Conv2D(filters[2], (1, 1), strides=(stride, stride), padding='valid')(x_shortcut)
x_shortcut = layers.BatchNormalization()(x_shortcut)
x = layers.Add()([x, x_shortcut])
x = layers.Activation('relu')(x)
return x
def build_resnet15(input_shape=(224, 224, 3), num_classes=5):
input_tensor = layers.Input(shape=input_shape)
x = layers.Conv2D(32, (7, 7), strides=(2, 2), padding='valid')(input_tensor)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.MaxPooling2D((3, 3), strides=(2, 2), padding='valid')(x)
x = conv_block(x, [32, 32, 64], stride=1)
x = identity_block(x, [64, 64, 64])
x = conv_block(x, [64, 64, 128], stride=2)
x = identity_block(x, [128, 128, 128])
x = conv_block(x, [128, 128, 256], stride=2)
x = identity_block(x, [256, 256, 256])
x = conv_block(x, [256, 256, 512], stride=2)
x = identity_block(x, [512, 512, 512])
x = layers.AveragePooling2D((4, 4), padding='valid')(x)
x = layers.Flatten()(x)
x = layers.Dense(512, activation='relu')(x)
x = layers.Dropout(0.2)(x)
output_tensor = layers.Dense(num_classes, activation='softmax')(x)
model = models.Model(inputs=input_tensor, outputs=output_tensor)
return model
# Instantiate the ResNet15 model
model = build_resnet15()
# Load the pre-trained weights
model.load_weights('model_weights.h5') # Replace with the actual path
# Define a function for making predictions
def predict(input_img):
# Preprocess the input image
input_img = input_img.resize((224, 224)) # Resize the image to match the model's input size
input_img_array = np.array(input_img) / 255.0 # Normalize pixel values
input_img_array = np.expand_dims(input_img_array, axis=0) # Add batch dimension
# Make predictions
predictions = model.predict(input_img_array)
predicted_class = np.argmax(predictions)
# Create a custom bar chart using gr.Tabs
prob_values = predictions[0].tolist()
bar_chart = gr.Bar(values=prob_values, labels=labels_list, title="Probabilities")
return input_img, {
"Predicted Class": labels_list[predicted_class],
"Probabilities": bar_chart
}
# Create a Gradio interface
gr.Interface(
predict,
inputs=gr.Image(label="Select an image", type="pil"),
outputs=[gr.Image(label="Processed Image"), gr.Label(label="Predicted Class"), gr.Tabs("Probabilities")],
title="ResNet15 Image Classification"
).launch()