app.py

import gradio as gr
import matplotlib.pyplot as plt
import numpy as np
import open3d as o3d
import os
from PIL import Image
import tempfile
import torch
from transformers import GLPNImageProcessor, GLPNForDepthEstimation


def predict_depth(image):
    # load and resize the input image
    new_height = 480 if image.height > 480 else image.height
    new_height -= (new_height % 32)
    new_width = int(new_height * image.width / image.height)
    diff = new_width % 32
    new_width = new_width - diff if diff < 16 else new_width + 32 - diff
    new_size = (new_width, new_height)
    image = image.resize(new_size)

    # prepare image for the model
    inputs = feature_extractor(images=image, return_tensors="pt")

    # get the prediction from the model
    with torch.no_grad():
        outputs = model(**inputs)
        predicted_depth = outputs.predicted_depth

    output = predicted_depth.squeeze().cpu().numpy() * 1000.0

    # remove borders
    pad = 16
    output = output[pad:-pad, pad:-pad]
    image = image.crop((pad, pad, image.width - pad, image.height - pad))

    return image, output


def generate_mesh(image, depth_image, quality):
    width, height = image.size

    # depth_image = (depth_map * 255 / np.max(depth_map)).astype('uint8')
    image = np.array(image)

    # create rgbd image
    depth_o3d = o3d.geometry.Image(depth_image)
    image_o3d = o3d.geometry.Image(image)
    rgbd_image = o3d.geometry.RGBDImage.create_from_color_and_depth(image_o3d, depth_o3d,
                                                                    convert_rgb_to_intensity=False)

    # camera settings
    camera_intrinsic = o3d.camera.PinholeCameraIntrinsic()
    camera_intrinsic.set_intrinsics(width, height, 500, 500, width / 2, height / 2)

    # create point cloud
    pcd = o3d.geometry.PointCloud.create_from_rgbd_image(rgbd_image, camera_intrinsic)

    # outliers removal
    cl, ind = pcd.remove_statistical_outlier(nb_neighbors=20, std_ratio=20.0)
    pcd = pcd.select_by_index(ind)

    # estimate normals
    pcd.estimate_normals()
    pcd.orient_normals_to_align_with_direction(orientation_reference=(0., 0., -1.))

    # surface reconstruction
    mesh = o3d.geometry.TriangleMesh.create_from_point_cloud_poisson(pcd, depth=quality, n_threads=1)[0]

    # rotate the mesh
    rotation = mesh.get_rotation_matrix_from_xyz((np.pi, np.pi, 0))
    mesh.rotate(rotation, center=(0, 0, 0))
    mesh.scale(256, center=(0, 0, 0))

    # save the mesh
    temp_name = next(tempfile._get_candidate_names()) + '.obj'
    o3d.io.write_triangle_mesh(temp_name, mesh)

    return temp_name


def predict(image, quality):
    image, depth_map = predict_depth(image)
    depth_image = (depth_map * 255 / np.max(depth_map)).astype('uint8')
    mesh_path = generate_mesh(image, depth_image, quality + 5)
    colormap = plt.get_cmap('plasma')
    depth_image = (colormap(depth_image) * 255).astype('uint8')
    depth_image = Image.fromarray(depth_image)

    return depth_image, mesh_path


feature_extractor = GLPNImageProcessor.from_pretrained("vinvino02/glpn-nyu")
model = GLPNForDepthEstimation.from_pretrained("vinvino02/glpn-nyu")


# GUI
title = 'Image2Mesh'
description = 'Demo based on my <a href="https://towardsdatascience.com/generate-a-3d-mesh-from-an-image-with-python' \
              '-12210c73e5cc">article</a>. This demo predicts the depth of an image and then generates the 3D mesh. ' \
              'Choosing a higher quality increases the time to generate the mesh. You can download the mesh by ' \
              'clicking the top-right button on the 3D viewer. '
examples = [[f'examples/{name}', 3] for name in sorted(os.listdir('examples'))]

# example image source:
# N. Silberman, D. Hoiem, P. Kohli, and Rob Fergus,
# Indoor Segmentation and Support Inference from RGBD Images (2012)

iface = gr.Interface(
    fn=predict,
    inputs=[
        gr.Image(type='pil', label='Input Image'),
        gr.Slider(1, 5, step=1, value=3, label='Mesh quality')
    ],
    outputs=[
        gr.Image(label='Depth'),
        gr.Model3D(label='3D Model', clear_color=[0.0, 0.0, 0.0, 0.0])
    ],
    examples=examples,
    allow_flagging='never',
    cache_examples=False,
    title=title,
    description=description
)
iface.launch()