def main():
try:
os.makedirs(args.output_directory)
except:
pass
last_file_number = args.initial_file_number + args.total_scenes // args.num_scenes_per_file - 1
initial_file_number = args.initial_file_number
if os.path.isdir(args.output_directory):
files = os.listdir(args.output_directory)
for name in files:
number = int(name.replace(".h5", ""))
if number > last_file_number:
continue
if number < args.initial_file_number:
continue
if number < initial_file_number:
continue
initial_file_number = number + 1
total_scenes_to_render = args.total_scenes - args.num_scenes_per_file * (
initial_file_number - args.initial_file_number)
assert args.num_scenes_per_file <= total_scenes_to_render
# Colors
colors = []
for n in range(args.num_colors):
hue = n / args.num_colors
saturation = 1
lightness = 1
red, green, blue = colorsys.hsv_to_rgb(hue, saturation, lightness)
colors.append(np.array((red, green, blue, 1)))
renderer = OffscreenRenderer(viewport_width=args.image_size,
viewport_height=args.image_size)
archiver = Archiver(
directory=args.output_directory,
num_scenes_per_file=args.num_scenes_per_file,
image_size=(args.image_size, args.image_size),
num_observations_per_scene=args.num_observations_per_scene,
initial_file_number=initial_file_number)
for scene_index in tqdm(range(total_scenes_to_render)):
scene = build_scene(floor_textures,
wall_textures,
fix_light_position=args.fix_light_position)
place_objects(scene,
colors,
objects,
max_num_objects=args.max_num_objects,
discrete_position=args.discrete_position,
rotate_object=args.rotate_object)
camera_distance = 4.5
camera = PerspectiveCamera(yfov=math.pi / 4)
camera_node = Node(camera=camera, translation=np.array([0, 1, 1]))
scene.add_node(camera_node)
scene_data = SceneData((args.image_size, args.image_size),
args.num_observations_per_scene)
for observation_index in range(args.num_observations_per_scene):
rand_position_xz = np.random.normal(size=2)
rand_position_xz = camera_distance * rand_position_xz / np.linalg.norm(
rand_position_xz)
# Compute yaw and pitch
camera_direction = np.array(
[rand_position_xz[0], 0, rand_position_xz[1]])
yaw, pitch = compute_yaw_and_pitch(camera_direction)
camera_node.rotation = genearte_camera_quaternion(yaw, pitch)
camera_position = np.array(
[rand_position_xz[0], 1, rand_position_xz[1]])
camera_node.translation = camera_position
# Rendering
flags = RenderFlags.SHADOWS_DIRECTIONAL
if args.anti_aliasing:
flags |= RenderFlags.ANTI_ALIASING
image = renderer.render(scene, flags=flags)[0]
scene_data.add(image, camera_position, math.cos(yaw),
math.sin(yaw), math.cos(pitch), math.sin(pitch))
if args.visualize:
plt.clf()
plt.imshow(image)
plt.pause(1e-10)
archiver.add(scene_data)
renderer.delete()
def main():
try:
os.makedirs(args.output_directory)
except:
pass
last_file_number = args.initial_file_number + args.total_scenes // args.num_scenes_per_file - 1
initial_file_number = args.initial_file_number
if os.path.isdir(args.output_directory):
files = os.listdir(args.output_directory)
for name in files:
number = int(name.replace(".h5", ""))
if number > last_file_number:
continue
if number < args.initial_file_number:
continue
if number < initial_file_number:
continue
initial_file_number = number + 1
total_scenes_to_render = args.total_scenes - args.num_scenes_per_file * (
initial_file_number - args.initial_file_number)
assert args.num_scenes_per_file <= total_scenes_to_render
# Set GPU device
rtx.set_device(args.gpu_device)
# Initialize colors
colors = []
for n in range(args.num_colors):
hue = n / args.num_colors
saturation = 1
lightness = 1
red, green, blue = colorsys.hsv_to_rgb(hue, saturation, lightness)
colors.append((red, green, blue, 1))
screen_width = args.image_size
screen_height = args.image_size
# Setting up a raytracer
rt_args = rtx.RayTracingArguments()
rt_args.num_rays_per_pixel = 1024
rt_args.max_bounce = 3
rt_args.supersampling_enabled = args.anti_aliasing
rt_args.next_event_estimation_enabled = True
rt_args.ambient_light_intensity = 0.1
cuda_args = rtx.CUDAKernelLaunchArguments()
cuda_args.num_threads = 64
cuda_args.num_rays_per_thread = 32
renderer = rtx.Renderer()
render_buffer = np.zeros((screen_height, screen_width, 3),
dtype=np.float32)
archiver = Archiver(
directory=args.output_directory,
num_scenes_per_file=args.num_scenes_per_file,
image_size=(args.image_size, args.image_size),
num_observations_per_scene=args.num_observations_per_scene,
initial_file_number=initial_file_number)
camera = rtx.PerspectiveCamera(fov_rad=math.pi / 3,
aspect_ratio=screen_width / screen_height)
camera_distance = 2
for _ in tqdm(range(total_scenes_to_render)):
scene = build_scene(floor_textures,
wall_textures,
fix_light_position=args.fix_light_position)
place_objects(scene,
colors,
max_num_objects=args.max_num_objects,
discrete_position=args.discrete_position,
rotate_object=args.rotate_object)
scene_data = SceneData((args.image_size, args.image_size),
args.num_observations_per_scene)
for _ in range(args.num_observations_per_scene):
# Sample camera position
rand_position_xz = np.random.normal(size=2)
rand_position_xz = camera_distance * rand_position_xz / np.linalg.norm(
rand_position_xz)
camera_position = np.array(
(rand_position_xz[0], wall_height / 2, rand_position_xz[1]))
center = np.array((0, wall_height / 2, 0))
# Compute yaw and pitch
camera_direction = camera_position - center
yaw, pitch = compute_yaw_and_pitch(camera_direction)
camera.look_at(tuple(camera_position), tuple(center), up=(0, 1, 0))
renderer.render(scene, camera, rt_args, cuda_args, render_buffer)
# Convert to sRGB
image = np.power(np.clip(render_buffer, 0, 1), 1.0 / 2.2)
image = np.uint8(image * 255)
image = cv2.bilateralFilter(image, 3, 25, 25)
scene_data.add(image, camera_position, math.cos(yaw),
math.sin(yaw), math.cos(pitch), math.sin(pitch))
if args.visualize:
plt.clf()
plt.imshow(image)
plt.pause(1e-10)
archiver.add(scene_data)
def main():
try:
os.makedirs(args.output_directory)
except:
pass
last_file_number = args.initial_file_number + args.total_scenes // args.num_scenes_per_file - 1
initial_file_number = args.initial_file_number
if os.path.isdir(args.output_directory):
files = os.listdir(args.output_directory)
for name in files:
number = int(name.replace(".h5", ""))
if number > last_file_number:
continue
if number < args.initial_file_number:
continue
if number < initial_file_number:
continue
initial_file_number = number + 1
total_scenes_to_render = args.total_scenes - args.num_scenes_per_file * (
initial_file_number - args.initial_file_number)
assert args.num_scenes_per_file <= total_scenes_to_render
# Initialize colors
color_candidates = []
for n in range(args.num_colors):
hue = n / args.num_colors
saturation = 1
lightness = 1
red, green, blue = colorsys.hsv_to_rgb(hue, saturation, lightness)
color_candidates.append((red, green, blue))
scene, cube_nodes = build_scene(args.num_cubes, color_candidates)
camera = OrthographicCamera(xmag=0.9, ymag=0.9)
camera_node = Node(camera=camera)
scene.add_node(camera_node)
renderer = OffscreenRenderer(
viewport_width=args.image_size, viewport_height=args.image_size)
archiver = Archiver(
directory=args.output_directory,
num_scenes_per_file=args.num_scenes_per_file,
image_size=(args.image_size, args.image_size),
num_observations_per_scene=args.num_observations_per_scene,
initial_file_number=initial_file_number)
for scene_index in tqdm(range(total_scenes_to_render)):
camera_distance = 2
scene_data = SceneData((args.image_size, args.image_size),
args.num_observations_per_scene)
for observation_index in range(args.num_observations_per_scene):
# Generate random point on a sphere
camera_position = np.random.normal(size=3)
camera_position = camera_distance * camera_position / np.linalg.norm(
camera_position)
# Compute yaw and pitch
yaw, pitch = compute_yaw_and_pitch(camera_position)
camera_node.rotation = genearte_camera_quaternion(yaw, pitch)
camera_node.translation = camera_position
# Rendering
flags = RenderFlags.SHADOWS_DIRECTIONAL
if args.anti_aliasing:
flags |= RenderFlags.ANTI_ALIASING
image = renderer.render(scene, flags=flags)[0]
scene_data.add(image, camera_position, math.cos(yaw),
math.sin(yaw), math.cos(pitch), math.sin(pitch))
if args.visualize:
plt.clf()
plt.imshow(image)
plt.pause(1e-10)
archiver.add(scene_data)
# Change cube color and position
update_cube_color_and_position(cube_nodes, color_candidates)
# Transfer changes to the vertex buffer on gpu
udpate_vertex_buffer(cube_nodes)
renderer.delete()
def main():
try:
os.makedirs(args.output_directory)
except:
pass
# Colors
colors = []
for n in range(args.num_colors):
hue = n / args.num_colors
saturation = 1
lightness = 1
red, green, blue = colorsys.hsv_to_rgb(hue, saturation, lightness)
colors.append(np.array((red, green, blue, 1)))
renderer = OffscreenRenderer(viewport_width=args.image_size,
viewport_height=args.image_size)
archiver = Archiver(
directory=args.output_directory,
total_scenes=args.total_scenes,
num_scenes_per_file=min(args.num_scenes_per_file, args.total_scenes),
image_size=(args.image_size, args.image_size),
num_observations_per_scene=args.num_observations_per_scene,
initial_file_number=args.initial_file_number)
for scene_index in tqdm(range(args.total_scenes)):
scene = build_scene(floor_textures,
wall_textures,
fix_light_position=args.fix_light_position)
place_objects(scene,
colors,
objects,
max_num_objects=args.max_num_objects,
discrete_position=args.discrete_position,
rotate_object=args.rotate_object)
camera_distance = 3
camera = PerspectiveCamera(yfov=math.pi / 4)
camera_node = Node(camera=camera, translation=np.array([0, 1, 1]))
scene.add_node(camera_node)
scene_data = SceneData((args.image_size, args.image_size),
args.num_observations_per_scene)
for observation_index in range(args.num_observations_per_scene):
# Sample camera position
rand_position_xz = np.random.uniform(-3, 3, size=2)
rand_lookat_xz = np.random.uniform(-6, 6, size=2)
camera_position = np.array(
[rand_position_xz[0], 1, rand_position_xz[1]])
# Compute yaw and pitch
camera_direction = rand_position_xz - rand_lookat_xz
camera_direction = np.array(
[camera_direction[0], 0, camera_direction[1]])
yaw, pitch = compute_yaw_and_pitch(camera_direction)
camera_node.rotation = genearte_camera_quaternion(yaw, pitch)
camera_node.translation = camera_position
# Rendering
flags = RenderFlags.SHADOWS_DIRECTIONAL
if args.anti_aliasing:
flags |= RenderFlags.ANTI_ALIASING
image = renderer.render(scene, flags=flags)[0]
scene_data.add(image, camera_position, math.cos(yaw),
math.sin(yaw), math.cos(pitch), math.sin(pitch))
if args.visualize:
plt.clf()
plt.imshow(image)
plt.pause(1e-10)
archiver.add(scene_data)
renderer.delete()
def main():
# Set GPU device
rtx.set_device(args.gpu_device)
# Initialize colors
color_array = []
for n in range(args.num_colors):
hue = n / (args.num_colors - 1)
saturation = 0.9
lightness = 1
red, green, blue = colorsys.hsv_to_rgb(hue, saturation, lightness)
color_array.append((red, green, blue, 1))
screen_width = args.image_size
screen_height = args.image_size
# Setting up a raytracer
rt_args = rtx.RayTracingArguments()
rt_args.num_rays_per_pixel = 512
rt_args.max_bounce = 2
rt_args.supersampling_enabled = args.anti_aliasing
rt_args.ambient_light_intensity = 0.05
cuda_args = rtx.CUDAKernelLaunchArguments()
cuda_args.num_threads = 64
cuda_args.num_rays_per_thread = 32
renderer = rtx.Renderer()
render_buffer = np.zeros((screen_height, screen_width, 3),
dtype=np.float32)
archiver = Archiver(
directory=args.output_directory,
total_scenes=args.total_scenes,
num_scenes_per_file=min(args.num_scenes_per_file, args.total_scenes),
image_size=(args.image_size, args.image_size),
num_observations_per_scene=args.num_observations_per_scene,
initial_file_number=args.initial_file_number)
camera = rtx.OrthographicCamera()
for _ in tqdm(range(args.total_scenes)):
scene = build_scene(args.num_cubes, color_array)
scene_data = SceneData((args.image_size, args.image_size),
args.num_observations_per_scene)
camera_distance = 1
for _ in range(args.num_observations_per_scene):
# Generate random point on a sphere
camera_position = np.random.normal(size=3)
camera_position = camera_distance * camera_position / np.linalg.norm(
camera_position)
# Compute yaw and pitch
yaw, pitch = compute_yaw_and_pitch(camera_position)
center = (0, 0, 0)
camera.look_at(tuple(camera_position), center, up=(0, 1, 0))
renderer.render(scene, camera, rt_args, cuda_args, render_buffer)
# Convert to sRGB
image = np.power(np.clip(render_buffer, 0, 1), 1.0 / 2.2)
image = np.uint8(image * 255)
image = cv2.bilateralFilter(image, 3, 25, 25)
if args.visualize:
plt.clf()
plt.imshow(image)
plt.pause(1e-10)
scene_data.add(image, camera_position, math.cos(yaw),
math.sin(yaw), math.cos(pitch), math.sin(pitch))
archiver.add(scene_data)
def main():
try:
os.makedirs(args.output_directory)
except:
pass
# Load MNIST images
mnist_images = load_mnist_images()
renderer = OffscreenRenderer(viewport_width=args.image_size,
viewport_height=args.image_size)
archiver = Archiver(
directory=args.output_directory,
total_scenes=args.total_scenes,
num_scenes_per_file=min(args.num_scenes_per_file, args.total_scenes),
image_size=(args.image_size, args.image_size),
num_observations_per_scene=args.num_observations_per_scene,
initial_file_number=args.initial_file_number)
for scene_index in tqdm(range(args.total_scenes)):
scene = build_scene(floor_textures,
wall_textures,
fix_light_position=args.fix_light_position)
place_dice(scene,
mnist_images,
discrete_position=args.discrete_position,
rotate_dice=args.rotate_dice)
camera_distance = 4
camera = PerspectiveCamera(yfov=math.pi / 4)
camera_node = Node(camera=camera, translation=np.array([0, 1, 1]))
scene.add_node(camera_node)
scene_data = SceneData((args.image_size, args.image_size),
args.num_observations_per_scene)
for observation_index in range(args.num_observations_per_scene):
rand_position_xz = np.random.normal(size=2)
rand_position_xz = camera_distance * rand_position_xz / np.linalg.norm(
rand_position_xz)
# Compute yaw and pitch
camera_direction = np.array(
[rand_position_xz[0], 0, rand_position_xz[1]])
yaw, pitch = compute_yaw_and_pitch(camera_direction)
camera_node.rotation = genearte_camera_quaternion(yaw, pitch)
camera_position = np.array(
[rand_position_xz[0], 1, rand_position_xz[1]])
camera_node.translation = camera_position
# Rendering
flags = RenderFlags.SHADOWS_DIRECTIONAL
if args.anti_aliasing:
flags |= RenderFlags.ANTI_ALIASING
image = renderer.render(scene, flags=flags)[0]
scene_data.add(image, camera_position, math.cos(yaw),
math.sin(yaw), math.cos(pitch), math.sin(pitch))
if args.visualize:
plt.clf()
plt.imshow(image)
plt.pause(1e-10)
archiver.add(scene_data)
renderer.delete()
def main():
try:
os.makedirs(args.output_directory)
except:
pass
# Load MNIST images
mnist_images = load_mnist_images()
# Set GPU device
rtx.set_device(args.gpu_device)
screen_width = args.image_size
screen_height = args.image_size
# Setting up a raytracer
rt_args = rtx.RayTracingArguments()
rt_args.num_rays_per_pixel = 1024
rt_args.max_bounce = 3
rt_args.supersampling_enabled = args.anti_aliasing
rt_args.next_event_estimation_enabled = True
rt_args.ambient_light_intensity = 0.1
cuda_args = rtx.CUDAKernelLaunchArguments()
cuda_args.num_threads = 64
cuda_args.num_rays_per_thread = 32
renderer = rtx.Renderer()
render_buffer = np.zeros((screen_height, screen_width, 3),
dtype=np.float32)
archiver = Archiver(
directory=args.output_directory,
total_scenes=args.total_scenes,
num_scenes_per_file=min(args.num_scenes_per_file, args.total_scenes),
image_size=(args.image_size, args.image_size),
num_observations_per_scene=args.num_observations_per_scene,
initial_file_number=args.initial_file_number)
camera = rtx.PerspectiveCamera(fov_rad=math.pi / 3,
aspect_ratio=screen_width / screen_height)
camera_distance = 2
for _ in tqdm(range(args.total_scenes)):
scene = build_scene(floor_textures,
wall_textures,
fix_light_position=args.fix_light_position)
place_dice(scene,
mnist_images,
discrete_position=args.discrete_position,
rotate_dice=args.rotate_dice)
scene_data = SceneData((args.image_size, args.image_size),
args.num_observations_per_scene)
for _ in range(args.num_observations_per_scene):
# Sample camera position
rand_position_xz = np.random.normal(size=2)
rand_position_xz = camera_distance * rand_position_xz / np.linalg.norm(
rand_position_xz)
camera_position = np.array(
(rand_position_xz[0], wall_height / 2, rand_position_xz[1]))
center = np.array((0, wall_height / 2, 0))
# Compute yaw and pitch
camera_direction = camera_position - center
yaw, pitch = compute_yaw_and_pitch(camera_direction)
camera.look_at(tuple(camera_position), tuple(center), up=(0, 1, 0))
renderer.render(scene, camera, rt_args, cuda_args, render_buffer)
# Convert to sRGB
image = np.power(np.clip(render_buffer, 0, 1), 1.0 / 2.2)
image = np.uint8(image * 255)
image = cv2.bilateralFilter(image, 3, 25, 25)
scene_data.add(image, camera_position, math.cos(yaw),
math.sin(yaw), math.cos(pitch), math.sin(pitch))
if args.visualize:
plt.clf()
plt.imshow(image)
plt.pause(1e-10)
archiver.add(scene_data)