import gradio as gr import torch from PIL import Image import numpy as np from diffusers import StableDiffusionDepth2ImgPipeline from pathlib import Path device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') dept2img = StableDiffusionDepth2ImgPipeline.from_pretrained( "stabilityai/stable-diffusion-2-depth", torch_dtype=torch.float16, ).to(device) def pad_image(input_image): pad_w, pad_h = np.max(((2, 2), np.ceil( np.array(input_image.size) / 64).astype(int)), axis=0) * 64 - input_image.size im_padded = Image.fromarray( np.pad(np.array(input_image), ((0, pad_h), (0, pad_w), (0, 0)), mode='edge')) w, h = im_padded.size if w == h: return im_padded elif w > h: new_image = Image.new(im_padded.mode, (w, w), (0, 0, 0)) new_image.paste(im_padded, (0, (w - h) // 2)) return new_image else: new_image = Image.new(im_padded.mode, (h, h), (0, 0, 0)) new_image.paste(im_padded, ((h - w) // 2, 0)) return new_image def predict(input_image, prompt, negative_prompt, steps, num_samples, scale, seed, strength, depth_image=None): depth = None if depth_image is not None: depth_image = pad_image(depth_image) depth_image = depth_image.resize((512, 512)) depth = np.array(depth_image.convert("L")) depth = depth.astype(np.float32) / 255.0 depth = depth[None, None] depth = torch.from_numpy(depth) init_image = input_image.convert("RGB") image = pad_image(init_image) # resize to integer multiple of 32 image = image.resize((512, 512)) result = dept2img( image=image, prompt=prompt, negative_prompt=negative_prompt, depth_image=depth, seed=seed, strength=strength, num_inference_steps=steps, guidance_scale=scale, num_images_per_prompt=num_samples, ) return result['images'] block = gr.Blocks().queue() with block: with gr.Row(): with gr.Column(): gr.Markdown("## Stable Diffusion 2 Depth2Img") gr.HTML("

Duplicate Space

") with gr.Row(): with gr.Column(): input_image = gr.Image(source='upload', type="pil") depth_image = gr.Image( source='upload', type="pil", label="Depth image Optional", value=None) prompt = gr.Textbox(label="Prompt") negative_prompt = gr.Textbox(label="Negative Pompt") run_button = gr.Button(label="Run") with gr.Accordion("Advanced options", open=False): num_samples = gr.Slider( label="Images", minimum=1, maximum=4, value=1, step=1) steps = gr.Slider(label="Steps", minimum=1, maximum=50, value=50, step=1) scale = gr.Slider( label="Guidance Scale", minimum=0.1, maximum=30.0, value=9.0, step=0.1 ) strength = gr.Slider( label="Strength", minimum=0.0, maximum=1.0, value=0.9, step=0.01 ) seed = gr.Slider( label="Seed", minimum=0, maximum=2147483647, step=1, randomize=True, ) with gr.Column(): gallery = gr.Gallery(label="Generated images", show_label=False).style( grid=[2], height="auto") gr.Examples( examples=[ ["./examples/baby.jpg", "high definition photo of a baby astronaut space walking at the international space station with earth seeing from above in the background", "", 50, 4, 9.0, 123123123, 0.8, None], ["./examples/gol.jpg", "professional photo of a Elmo jumping between two high rises, beautiful colorful city landscape in the background", "", 50, 4, 9.0, 1734133747, 0.9, None], ["./examples/bag.jpg", "a photo of a bag of cookies in the bathroom", "low light, dark, blurry", 50, 4, 9.0, 1734133747, 0.9, "./examples/depth.jpg"], ["./examples/smile_face.jpg", "a hand holding a very spherical orange", "low light, dark, blurry", 50, 4, 6.0, 961736534, 0.5, "./examples/smile_depth.jpg"] ], inputs=[input_image, prompt, negative_prompt, steps, num_samples, scale, seed, strength, depth_image], outputs=[gallery], fn=predict, cache_examples=True, ) run_button.click(fn=predict, inputs=[input_image, prompt, negative_prompt, steps, num_samples, scale, seed, strength, depth_image], outputs=[gallery]) block.launch(show_api=False)