def __init__(self, input_obj_file, input_ref_file):
super(Model, self).__init__()
# Load the OBJ file.
vertices, faces = neural_renderer.load_obj(input_obj_file)
self.vertices = torch.as_tensor(vertices[None, :, :])
self.faces = torch.as_tensor(faces)
# Create the textures.
vertices_t, faces_t, textures = neural_renderer.create_textures(
self.faces.shape[0], texture_size=4)
self.vertices_t = torch.as_tensor(vertices_t[None, :, :])
self.faces_t = torch.as_tensor(faces_t)
self.textures = torch.nn.Parameter(torch.as_tensor(
textures[None, :, :, :]),
requires_grad=True)
# Load the reference image.
self.image_ref = neural_renderer.imread(input_ref_file)
self.image_ref = torch.as_tensor(self.image_ref)
# Set up the renderer.
renderer = neural_renderer.Renderer()
renderer.perspective = False
self.renderer = renderer
def __init__(self, filename_obj, filename_ref):
super(Model, self).__init__()
# load .obj
vertices, faces = neural_renderer.load_obj(filename_obj)
self.vertices = torch.as_tensor(vertices[None, :, :])
self.faces = torch.as_tensor(faces)
# create textures
vertices_t, faces_t, textures = neural_renderer.create_textures(
self.faces.shape[0], texture_size=4)
self.vertices_t = torch.as_tensor(vertices_t[None, :, :])
self.faces_t = torch.as_tensor(faces_t)
self.textures = torch.nn.Parameter(torch.as_tensor(
textures[None, :, :, :]),
requires_grad=True)
# load reference image
self.image_ref = neural_renderer.imread(filename_ref)
self.image_ref = torch.as_tensor(self.image_ref)
# setup renderer
renderer = neural_renderer.Renderer()
renderer.perspective = False
self.renderer = renderer
def test_backward_case1(self):
vertices = [
[0.1, 0.1, 1.],
[-0.1, 0.1, 1.],
[-0.1, -0.1, 1.],
[0.1, -0.1, 1.],
]
faces = [[0, 1, 2], [0, 2, 3]]
ref = neural_renderer_torch.imread('./tests_torch/data/gradient.png')
ref = 1 - ref
ref = ref[:, :, 0]
ref = torch.as_tensor(ref).cuda()
vertices = np.array(vertices, 'float32')
faces = np.array(faces, 'int32')
vertices, faces = neural_renderer_torch.to_gpu((vertices, faces))
vertices = Parameter(vertices, True)
faces = torch.as_tensor(faces)
optimizer = torch.optim.Adam([vertices], lr=0.005)
ref = torch.as_tensor(ref)
os.makedirs("tmp", exist_ok=True)
for i in range(350):
params = RasterizeParam()
hyperparams = RasterizeHyperparam(image_size=256,
anti_aliasing=False)
images = neural_renderer_torch.rasterize_silhouettes(
vertices[None, :, :], faces, params, hyperparams)
image = images[0]
iou = torch.sum(image * ref) / torch.sum(image + ref - image * ref)
iou = torch.as_tensor(1, dtype=torch.float32,
device=iou.device) - iou
loss = iou
optimizer.zero_grad()
loss.backward()
optimizer.step()
pylab.imsave('tmp/t%d.png' % i, image.data.cpu().numpy())
print(i, loss.data, iou.data)
if float(iou.data) < 0.01:
return
raise Exception
def test_case1(self):
data = [
[
'./tests_chainer/data/4e49873292196f02574b5684eaec43e9/model.obj',
neural_renderer_torch.get_points_from_angles(2.5, 10, -90),
'./tests_chainer/data/4e49873292196f02574b5684eaec43e9.png',
],
[
'./tests_chainer/data/1cde62b063e14777c9152a706245d48/model.obj',
neural_renderer_torch.get_points_from_angles(2.5, 10, 60),
'./tests_chainer/data/1cde62b063e14777c9152a706245d48.png',
]
]
filename_tmp = './tests_chainer/data/tmp.obj'
renderer = neural_renderer_torch.Renderer()
renderer.draw_backside = False
for i, (filename, viewpoint, reference) in enumerate(data):
renderer.viewpoints = viewpoint
ref = neural_renderer_torch.imread(reference)
vertices, faces, vertices_t, faces_t, textures = neural_renderer_torch.load_obj(
filename, load_textures=True)
neural_renderer_torch.save_obj(filename_tmp, vertices, faces,
vertices_t, faces_t, textures)
vertices, faces, vertices_t, faces_t, textures = neural_renderer_torch.load_obj(
filename_tmp, load_textures=True)
vertices, faces, vertices_t, faces_t, textures = neural_renderer_torch.to_gpu(
(vertices[None, :, :], faces, vertices_t[None, :, :], faces_t,
textures[None, :, :, :]))
images = renderer.render(vertices, faces, vertices_t, faces_t,
textures).data
image = images[0].transpose((1, 2, 0))
chainer.testing.assert_allclose(ref, image, atol=1e-2, rtol=1e-2)
for f in glob.glob('./tests_chainer/data/tmp*'):
os.remove(f)
def __init__(self, input_obj_file, input_ref_file=None):
super(Model, self).__init__()
# Load the OBJ file.
vertices, faces = neural_renderer_torch.load_obj(input_obj_file)
self.vertices = torch.as_tensor(vertices[None, :, :])
self.faces = torch.as_tensor(faces)
# Load the reference image.
if input_ref_file is not None:
self.image_ref = torch.as_tensor(
neural_renderer_torch.imread(input_ref_file))
else:
self.image_ref = None
# Set up the camera parameters.
self.camera_position = torch.nn.Parameter(
torch.tensor([6, 10, -14], dtype=torch.float32))
# Set up the renderer.
renderer = neural_renderer_torch.Renderer()
renderer.viewpoints = self.camera_position
self.renderer = renderer