app.py

import os
from PIL import Image
import torch

from point_e.diffusion.configs import DIFFUSION_CONFIGS, diffusion_from_config
from point_e.diffusion.sampler import PointCloudSampler
from point_e.models.download import load_checkpoint
from point_e.models.configs import MODEL_CONFIGS, model_from_config
from point_e.util.plotting import plot_point_cloud
from point_e.util.pc_to_mesh import marching_cubes_mesh

import skimage.measure

from pyntcloud import PyntCloud
import matplotlib.colors
import plotly.graph_objs as go

import trimesh

import gradio as gr


state = ""
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

def set_state(s):
    print(s)
    global state
    state = s

def get_state():
    return state

set_state('Creating txt2mesh model...')
t2m_name = 'base40M-textvec'
t2m_model = model_from_config(MODEL_CONFIGS[t2m_name], device)
t2m_model.eval()
base_diffusion_t2m = diffusion_from_config(DIFFUSION_CONFIGS[t2m_name])

set_state('Downloading txt2mesh checkpoint...')
t2m_model.load_state_dict(load_checkpoint(t2m_name, device))


def load_img2mesh_model(model_name):
    set_state(f'Creating img2mesh model {model_name}...')
    i2m_name = model_name
    i2m_model = model_from_config(MODEL_CONFIGS[i2m_name], device)
    i2m_model.eval()
    base_diffusion_i2m = diffusion_from_config(DIFFUSION_CONFIGS[i2m_name])

    set_state(f'Downloading img2mesh checkpoint {model_name}...')
    i2m_model.load_state_dict(load_checkpoint(i2m_name, device))

    return i2m_model, base_diffusion_i2m

img2mesh_model_name = 'base40M' #'base300M' #'base1B'
i2m_model, base_diffusion_i2m = load_img2mesh_model(img2mesh_model_name)


set_state('Creating upsample model...')
upsampler_model = model_from_config(MODEL_CONFIGS['upsample'], device)
upsampler_model.eval()
upsampler_diffusion = diffusion_from_config(DIFFUSION_CONFIGS['upsample'])

set_state('Downloading upsampler checkpoint...')
upsampler_model.load_state_dict(load_checkpoint('upsample', device))

set_state('Creating SDF model...')
sdf_name = 'sdf'
sdf_model = model_from_config(MODEL_CONFIGS[sdf_name], device)
sdf_model.eval()

set_state('Loading SDF model...')
sdf_model.load_state_dict(load_checkpoint(sdf_name, device))

stable_diffusion = gr.Blocks.load(name="spaces/runwayml/stable-diffusion-v1-5")


set_state('')

def get_sampler(model_name, txt2obj, guidance_scale):

    global img2mesh_model_name
    global base_diffusion_i2m
    global i2m_model
    if model_name != img2mesh_model_name:
        img2mesh_model_name = model_name
        i2m_model, base_diffusion_i2m = load_img2mesh_model(model_name)

    return PointCloudSampler(
            device=device,
            models=[t2m_model if txt2obj else i2m_model, upsampler_model],
            diffusions=[base_diffusion_t2m if txt2obj else base_diffusion_i2m, upsampler_diffusion],
            num_points=[1024, 4096 - 1024],
            aux_channels=['R', 'G', 'B'],
            guidance_scale=[guidance_scale, 0.0 if txt2obj else guidance_scale],
            model_kwargs_key_filter=('texts', '') if txt2obj else ("*",)
        )

def generate_txt2img(prompt):

    prompt = f"“a 3d rendering of {prompt}, full view, white background"
    gallery_dir = stable_diffusion(prompt, fn_index=2)
    imgs = [os.path.join(gallery_dir, img) for img in os.listdir(gallery_dir) if os.path.splitext(img)[1] == '.jpg']

    return imgs[0], gr.update(visible=True)

def generate_3D(input, model_name='base40M', guidance_scale=3.0, grid_size=32):

    set_state('Entered generate function...')

    if isinstance(input, Image.Image):
        input = prepare_img(input)

    # if input is a string, it's a text prompt
    sampler = get_sampler(model_name, txt2obj=True if isinstance(input, str) else False, guidance_scale=guidance_scale)

    # Produce a sample from the model.
    set_state('Sampling...')
    samples = None
    kw_args = dict(texts=[input]) if isinstance(input, str) else dict(images=[input])
    for x in sampler.sample_batch_progressive(batch_size=1, model_kwargs=kw_args):
        samples = x

    set_state('Converting to point cloud...')
    pc = sampler.output_to_point_clouds(samples)[0]

    set_state('Saving point cloud...')
    with open("point_cloud.ply", "wb") as f:
        pc.write_ply(f)

    set_state('Converting to mesh...')
    save_ply(pc, 'mesh.ply', grid_size)

    set_state('')

    return pc_to_plot(pc), ply_to_obj('mesh.ply', '3d_model.obj'), gr.update(value=['3d_model.obj', 'mesh.ply', 'point_cloud.ply'], visible=True)

def prepare_img(img):

    w, h = img.size
    if w > h:
        img = img.crop((w - h) / 2, 0, w - (w - h) / 2, h)
    else:
        img = img.crop((0, (h - w) / 2, w, h - (h - w) / 2))

    # resize to 256x256
    img = img.resize((256, 256))
    
    return img

def pc_to_plot(pc):

    return go.Figure(
        data=[
            go.Scatter3d(
                x=pc.coords[:,0], y=pc.coords[:,1], z=pc.coords[:,2], 
                mode='markers',
                marker=dict(
                  size=2,
                  color=['rgb({},{},{})'.format(r,g,b) for r,g,b in zip(pc.channels["R"], pc.channels["G"], pc.channels["B"])],
              )
            )
        ],
        layout=dict(
            scene=dict(xaxis=dict(visible=False), yaxis=dict(visible=False), zaxis=dict(visible=False))
        ),
    )

def ply_to_obj(ply_file, obj_file):
    mesh = trimesh.load(ply_file)
    mesh.export(obj_file)

    return obj_file

def save_ply(pc, file_name, grid_size):

    # Produce a mesh (with vertex colors)
    mesh = marching_cubes_mesh(
        pc=pc,
        model=sdf_model,
        batch_size=4096,
        grid_size=grid_size, # increase to 128 for resolution used in evals
        fill_vertex_channels=True,
        progress=True,
    )

    # Write the mesh to a PLY file to import into some other program.
    with open(file_name, 'wb') as f:
        mesh.write_ply(f)


with gr.Blocks() as app:
    gr.Markdown("## Point-E text-to-3D Demo")
    gr.Markdown("This is a demo for [Point-E: A System for Generating 3D Point Clouds from Complex Prompts](https://arxiv.org/abs/2212.08751) by OpenAI. Check out the [GitHub repo](https://github.com/openai/point-e) for more information.")
    gr.HTML("""To skip the queue you can duplicate this space:
            <br><a href="https://huggingface.co/spaces/anzorq/point-e_demo?duplicate=true"><img src="https://img.shields.io/badge/-Duplicate%20Space-blue?labelColor=white&style=flat&logo=data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABAAAAAQCAYAAAAf8/9hAAAAAXNSR0IArs4c6QAAAP5JREFUOE+lk7FqAkEURY+ltunEgFXS2sZGIbXfEPdLlnxJyDdYB62sbbUKpLbVNhyYFzbrrA74YJlh9r079973psed0cvUD4A+4HoCjsA85X0Dfn/RBLBgBDxnQPfAEJgBY+A9gALA4tcbamSzS4xq4FOQAJgCDwV2CPKV8tZAJcAjMMkUe1vX+U+SMhfAJEHasQIWmXNN3abzDwHUrgcRGmYcgKe0bxrblHEB4E/pndMazNpSZGcsZdBlYJcEL9Afo75molJyM2FxmPgmgPqlWNLGfwZGG6UiyEvLzHYDmoPkDDiNm9JR9uboiONcBXrpY1qmgs21x1QwyZcpvxt9NS09PlsPAAAAAElFTkSuQmCC&logoWidth=14" alt="Duplicate Space"></a>
            <br>Don't forget to change space hardware to <b>GPU</b> after duplicating it.""")
    
    with gr.Row():
        with gr.Column():
            with gr.Tab("Text to 3D"):
                prompt = gr.Textbox(label="Prompt", placeholder="A cactus in a pot")
                btn_generate_txt2obj = gr.Button(value="Generate")
            
            with gr.Tab("Image to 3D"):
                img = gr.Image(label="Image")
                gr.Markdown("Best results with images of 3D objects with no shadows on a white background.")
                btn_generate_img2obj = gr.Button(value="Generate")

            with gr.Tab("Text to Image to 3D"):
                gr.Markdown("Generate an image with Stable Diffusion, then convert it to 3D. Just enter the object you want to generate.")
                prompt_sd = gr.Textbox(label="Prompt", placeholder="a 3d rendering of [your prompt], full view, white background")
                btn_generate_txt2sd = gr.Button(value="Generate image")
                img_sd = gr.Image(label="Image")
                btn_generate_sd2obj = gr.Button(value="Convert to 3D", visible=False)

            with gr.Accordion("Advanced settings", open=False):
                dropdown_models = gr.Dropdown(label="Model", value="base40M", choices=["base40M", "base300M"]) #, "base1B"])
                guidance_scale = gr.Slider(label="Guidance scale", value=3.0, minimum=3.0, maximum=10.0, step=0.1)
                grid_size = gr.Slider(label="Grid size (for .obj 3D model)", value=32, minimum=16, maximum=128, step=16)

        with gr.Column():
            plot = gr.Plot(label="Point cloud")
            # btn_pc_to_obj = gr.Button(value="Convert to OBJ", visible=False)
            model_3d = gr.Model3D(value=None)
            file_out = gr.File(label="Files", visible=False)
            
            # state_info = state_info = gr.Textbox(label="State", show_label=False).style(container=False)


        # inputs = [dropdown_models, prompt, img, guidance_scale, grid_size]
        outputs = [plot, model_3d, file_out]

        prompt.submit(generate_3D, inputs=[prompt, dropdown_models, guidance_scale, grid_size], outputs=outputs)
        btn_generate_txt2obj.click(generate_3D, inputs=[prompt, dropdown_models, guidance_scale, grid_size], outputs=outputs, api_name="generate_txt2obj")
        
        btn_generate_img2obj.click(generate_3D, inputs=[img, dropdown_models, guidance_scale, grid_size], outputs=outputs, api_name="generate_img2obj")

        prompt_sd.submit(generate_txt2img, inputs=prompt_sd, outputs=[img_sd, btn_generate_sd2obj])
        btn_generate_txt2sd.click(generate_txt2img, inputs=prompt_sd, outputs=[img_sd, btn_generate_sd2obj], queue=False)
        btn_generate_sd2obj.click(generate_3D, inputs=[img, dropdown_models, guidance_scale, grid_size], outputs=outputs)

        # btn_pc_to_obj.click(ply_to_obj, inputs=plot, outputs=[model_3d, file_out])

    gr.Examples(
        examples=[
            ["a cactus in a pot"],
            ["a round table with floral tablecloth"],
            ["a red kettle"],
            ["a vase with flowers"],
            ["a sports car"],
            ["a man"],
        ],
        inputs=[prompt],
        outputs=outputs,
        fn=generate_3D,
        cache_examples=True
    )

    gr.Examples(
        examples=[
            ["images/corgi.png"],
            ["images/cube_stack.jpg"],
            ["images/chair.png"],
        ],
        inputs=[img],
        outputs=outputs,
        fn=generate_3D,
        cache_examples=True
    )

    # app.load(get_state, inputs=[], outputs=state_info, every=0.5, show_progress=False)

    gr.HTML("""
    <br><br>
    <div style="border-top: 1px solid #303030;">
      <br>
      <p>Space by:<br>
      <a href="https://twitter.com/hahahahohohe"><img src="https://img.shields.io/twitter/follow/hahahahohohe?label=%40anzorq&style=social" alt="Twitter Follow"></a><br>
      <a href="https://github.com/qunash"><img alt="GitHub followers" src="https://img.shields.io/github/followers/qunash?style=social" alt="Github Follow"></a></p><br>
      <a href="https://www.buymeacoffee.com/anzorq" target="_blank"><img src="https://cdn.buymeacoffee.com/buttons/v2/default-yellow.png" alt="Buy Me A Coffee" style="height: 30px !important;width: 102px !important;" ></a><br><br>
      <p><img src="https://visitor-badge.glitch.me/badge?page_id=anzorq.point-e_demo" alt="visitors"></p>
    </div>
    """)

app.queue(max_size=250, concurrency_count=6).launch()