M3000j's picture
Upload folder using huggingface_hub
31726e5 verified
import pydiffvg
import torch
import skimage
import numpy as np
# Use GPU if available
pydiffvg.set_use_gpu(torch.cuda.is_available())
canvas_width, canvas_height = 256, 256
ellipse = pydiffvg.Ellipse(radius = torch.tensor([60.0, 30.0]),
center = torch.tensor([128.0, 128.0]))
shapes = [ellipse]
ellipse_group = pydiffvg.ShapeGroup(shape_ids = torch.tensor([0]),
fill_color = torch.tensor([0.3, 0.6, 0.3, 1.0]))
shape_groups = [ellipse_group]
scene_args = pydiffvg.RenderFunction.serialize_scene(\
canvas_width, canvas_height, shapes, shape_groups)
render = pydiffvg.RenderFunction.apply
img = render(256, # width
256, # height
2, # num_samples_x
2, # num_samples_y
0, # seed
None, # background_image
*scene_args)
# The output image is in linear RGB space. Do Gamma correction before saving the image.
pydiffvg.imwrite(img.cpu(), 'results/single_ellipse/target.png', gamma=2.2)
target = img.clone()
# Move the ellipse to produce initial guess
# normalize radius & center for easier learning rate
radius_n = torch.tensor([20.0 / 256.0, 40.0 / 256.0], requires_grad=True)
center_n = torch.tensor([108.0 / 256.0, 138.0 / 256.0], requires_grad=True)
color = torch.tensor([0.3, 0.2, 0.8, 1.0], requires_grad=True)
ellipse.radius = radius_n * 256
ellipse.center = center_n * 256
ellipse_group.fill_color = color
scene_args = pydiffvg.RenderFunction.serialize_scene(\
canvas_width, canvas_height, shapes, shape_groups)
img = render(256, # width
256, # height
2, # num_samples_x
2, # num_samples_y
1, # seed
None, # background_image
*scene_args)
pydiffvg.imwrite(img.cpu(), 'results/single_ellipse/init.png', gamma=2.2)
# Optimize for radius & center
optimizer = torch.optim.Adam([radius_n, center_n, color], lr=1e-2)
# Run 50 Adam iterations.
for t in range(50):
print('iteration:', t)
optimizer.zero_grad()
# Forward pass: render the image.
ellipse.radius = radius_n * 256
ellipse.center = center_n * 256
ellipse_group.fill_color = color
scene_args = pydiffvg.RenderFunction.serialize_scene(\
canvas_width, canvas_height, shapes, shape_groups)
img = render(256, # width
256, # height
2, # num_samples_x
2, # num_samples_y
t+1, # seed
None, # background_image
*scene_args)
# Save the intermediate render.
pydiffvg.imwrite(img.cpu(), 'results/single_ellipse/iter_{}.png'.format(t), gamma=2.2)
# Compute the loss function. Here it is L2.
loss = (img - target).pow(2).sum()
print('loss:', loss.item())
# Backpropagate the gradients.
loss.backward()
# Print the gradients
print('radius.grad:', radius_n.grad)
print('center.grad:', center_n.grad)
print('color.grad:', color.grad)
# Take a gradient descent step.
optimizer.step()
# Print the current params.
print('radius:', ellipse.radius)
print('center:', ellipse.center)
print('color:', ellipse_group.fill_color)
# Render the final result.
ellipse.radius = radius_n * 256
ellipse.center = center_n * 256
ellipse_group.fill_color = color
scene_args = pydiffvg.RenderFunction.serialize_scene(\
canvas_width, canvas_height, shapes, shape_groups)
img = render(256, # width
256, # height
2, # num_samples_x
2, # num_samples_y
52, # seed
None, # background_image
*scene_args)
# Save the images and differences.
pydiffvg.imwrite(img.cpu(), 'results/single_ellipse/final.png')
# Convert the intermediate renderings to a video.
from subprocess import call
call(["ffmpeg", "-framerate", "24", "-i",
"results/single_ellipse/iter_%d.png", "-vb", "20M",
"results/single_ellipse/out.mp4"])