Spaces:
Sleeping
Sleeping
| import gradio as gr | |
| from train import main | |
| import spaces | |
| import torch | |
| import tensorflow as tf | |
| # Function to calculate the aspect ratio | |
| def calculate_aspect_ratio(width, height): | |
| return width / height | |
| # Function to update image dimensions based on aspect ratio | |
| def update_dimensions(value, aspect_ratio, dim): | |
| if dim == "width": | |
| new_height = int(value / aspect_ratio) | |
| return gr.update(value=value, visible=True), gr.update(value=new_height, visible=True) | |
| elif dim == "height": | |
| new_width = int(value * aspect_ratio) | |
| return gr.update(value=new_width, visible=True), gr.update(value=value, visible=True) | |
| # Function to handle image size selection | |
| def handle_image_size_selection(img_size, content_img): | |
| if img_size == "custom size": | |
| height, width, _ = content_img.shape | |
| aspect_ratio = calculate_aspect_ratio(width, height) | |
| return (gr.update(value=width, visible=True), | |
| gr.update(value=height, visible=True), | |
| gr.update(visible=True), | |
| gr.update(visible=True), | |
| aspect_ratio) | |
| else: | |
| return (gr.update(value=450, visible=False), | |
| gr.update(value=300, visible=False), | |
| gr.update(visible=False), | |
| gr.update(visible=False), | |
| None) | |
| # Define the function to process images | |
| def process_images(content_img, style_img, epochs, steps_per_epoch, learning_rate, content_loss_factor, style_loss_factor, img_size, img_width, img_height): | |
| print("Start processing") | |
| print(tf.config.list_physical_devices('GPU')) | |
| print(torch.cuda.get_device_name(0)) | |
| with tf.device("/physical_device:GPU:0"): | |
| output_img = main(content_img, style_img, epochs, steps_per_epoch, learning_rate, content_loss_factor, style_loss_factor, img_size, img_width, img_height) | |
| return output_img | |
| # @spaces.GPU | |
| def create_app(): | |
| with gr.Blocks() as demo: | |
| aspect_ratio = gr.State(None) | |
| with gr.Row(): | |
| with gr.Column(scale=1): | |
| content_img = gr.Image(type="numpy", label="Content Image") | |
| style_img = gr.Image(type="numpy", label="Style Image") | |
| process_button = gr.Button("Process") | |
| with gr.Accordion("Parameters", open=False): | |
| epochs = gr.Slider(minimum=1, maximum=100, step=1, label="Epochs", value=20) | |
| steps_per_epoch = gr.Slider(minimum=1, maximum=1000, step=1, label="Steps per Epoch", value=100) | |
| learning_rate = gr.Slider(minimum=0.0001, maximum=0.1, step=0.0001, label="Learning Rate", value=0.01) | |
| content_loss_factor = gr.Slider(minimum=0.1, maximum=10, step=0.1, label="Content Loss Factor", value=1.0) | |
| style_loss_factor = gr.Slider(minimum=0.1, maximum=1000, step=0.1, label="Style Loss Factor", value=100.0) | |
| img_size = gr.Dropdown(choices=["default size", "custom size"], label="Image Size", value="default size") | |
| img_width = gr.Number(label="Image Width", value=450, visible=False) | |
| img_height = gr.Number(label="Image Height", value=300, visible=False) | |
| with gr.Column(scale=1): | |
| output_img = gr.Image(label="Output Image") | |
| img_size.change( | |
| fn=handle_image_size_selection, | |
| inputs=[img_size, content_img], | |
| outputs=[img_width, img_height, img_width, img_height, aspect_ratio] | |
| ) | |
| img_width.change( | |
| fn=lambda w, ar: update_dimensions(w, ar, "width"), | |
| inputs=[img_width, aspect_ratio], | |
| outputs=[img_width, img_height] | |
| ) | |
| img_height.change( | |
| fn=lambda h, ar: update_dimensions(h, ar, "height"), | |
| inputs=[img_height, aspect_ratio], | |
| outputs=[img_width, img_height] | |
| ) | |
| process_button.click( | |
| process_images, | |
| inputs=[content_img, style_img, epochs, steps_per_epoch, learning_rate, content_loss_factor, style_loss_factor, img_size, img_width, img_height], | |
| outputs=[output_img] | |
| ) | |
| demo.launch() | |
| if __name__ == "__main__": | |
| create_app() | |