(
torch.ones(
1, faces_gt.shape[1], 2, 1, dtype=torch.float32,
device=device),
torch.zeros(
1, faces_gt.shape[1], 2, 1, dtype=torch.float32,
device=device),
torch.zeros(
1, faces_gt.shape[1], 2, 1, dtype=torch.float32,
device=device),
),
dim=-1,
)
# Initialize the renderer.
renderer = SoftRenderer(camera_mode="look_at", device=device)
camera_distance = 8.0
elevation = 0.0
azimuth = 0.0
renderer.set_eye_from_angles(camera_distance, elevation, azimuth)
pendulum_gt = SimplePendulum(args.mass, args.radius, args.gravity,
args.length, args.damping).to(device)
theta_init = torch.tensor([0.0, 3.0], device=device)
times = torch.arange(
args.starttime,
args.starttime + args.simsteps * args.dtime,
args.dtime,
device=device,
)
numsteps = times.numel()
# # writer = imageio.get_writer("cache/a.gif", mode="I")
# for i in trange(60):
# img = seqs[0, i]
# # writer.append_data(img.astype(np.uint8))
# imgs_gt.append(img.astype(np.uint8))
# # writer.close()
# Initialize the renderer
image_size = 256
camera_mode = "look_at"
camera_distance = 8.0
elevation = 30.0
azimuth = 0.0
# Initialize the renderer.
renderer = SoftRenderer(image_size=image_size,
camera_mode=camera_mode,
device=device)
renderer.set_eye_from_angles(camera_distance, elevation, azimuth)
sim_duration = 2.0 # seconds
fps = 30 # frames per second
sim_steps = int((sim_duration * fps) / 2)
# obj = get_primitive_obj(shape[0])
obj = get_primitive_obj(INT_TO_PRIMITIVE[shape[0]])
# print(obj)
mesh = TriangleMesh.from_obj(obj)
vertices = ((
meshutils.normalize_vertices(mesh.vertices) # + \
# torch.from_numpy(init_pos[i]).float().unsqueeze(0)
).to(device).unsqueeze(0))
model.contact_ke = 1.0e4
model.contact_kd = 1000.0
model.contact_kf = 1000.0
model.contact_mu = 0.5
model.particle_radius = 0.01
model.ground = False
target = torch.tensor((3.5, 0.0, 0.0))
initial_velocity = torch.tensor((1.0, 0.0, 0.0), requires_grad=True)
integrator = df.sim.SemiImplicitIntegrator()
# Setup SoftRasterizer
device = "cuda:0"
renderer = SoftRenderer(camera_mode="look_at", device=device)
camera_distance = 15.0
elevation = 0.0
azimuth = 0.0
renderer.set_eye_from_angles(camera_distance, elevation, azimuth)
faces = model.tri_indices
textures = torch.cat(
(
torch.zeros(1, faces.shape[-2], 2, 1, dtype=torch.float32, device=device),
torch.ones(1, faces.shape[-2], 2, 1, dtype=torch.float32, device=device),
torch.zeros(1, faces.shape[-2], 2, 1, dtype=torch.float32, device=device),
),
dim=-1,
)
# copy_scripts(
# os.path.abspath(
# os.path.join(os.path.dirname(os.path.abspath(__file__)), "../gradsim")
# ),
# os.path.join(args.out_dir, "gradsim"),
# )
shutil.copy(os.path.abspath(__file__), args.out_dir)
os.makedirs(os.path.join(args.out_dir, "obj"), exist_ok=True)
for obj in glob.glob(os.path.join(args.data_dir, "*.obj")):
shutil.copy(obj, os.path.join(args.out_dir, "obj"))
with open(os.path.join(args.out_dir, "args.yaml"), "w") as f:
yaml.dump(vars(args), f, default_flow_style=False)
# Initialize the renderer.
renderer = SoftRenderer(image_size=args.image_size,
camera_mode=args.camera_mode,
device=device)
renderer.set_eye_from_angles(args.camera_distance, args.elevation,
args.azimuth)
# Initialize sampler
prim_gen = PrimitiveGenerator(mass_scaling=args.mass_scaling,
mass_offset=args.mass_offset,
data_type=args.data_type,
data_dir=args.data_dir)
objects = sorted(glob.glob(os.path.join(args.data_dir, "*.obj")))
pos, ornt, linvel, angvel = None, None, None, None
try:
mesh=mesh,
scale=scale_gt,
density=density_gt,
ke=ke_gt,
kd=kd_gt,
kf=kf_gt,
mu=mu_gt,
)
model_gt = builder_gt.finalize("cpu")
model_gt.ground = True
integrator = df.sim.SemiImplicitIntegrator()
# Setup SoftRasterizer
device = "cuda:0"
renderer = SoftRenderer(camera_mode="look_at", device=device)
camera_distance = 10.0
elevation = 30.0
azimuth = 0.0
renderer.set_eye_from_angles(camera_distance, elevation, azimuth)
render_every = 60 * 4
vertices = torch.from_numpy(np.array(points).astype(np.float32))
faces = torch.from_numpy(np.asarray(indices).reshape((-1, 3)))
textures = torch.cat(
(
torch.ones(
1, faces.shape[-2], 2, 1, dtype=torch.float32, device=device),
torch.zeros(
1, faces.shape[-2], 2, 1, dtype=torch.float32, device=device),
# Add this force to the body.
body.add_external_force(gravity)
sim_duration = 2.0
fps = 30
sim_substeps = 32
dtime = (1 / 30) / sim_substeps
sim_steps = int(sim_duration / dtime)
render_every = sim_substeps
# Initialize the simulator with the body at the origin.
sim = Simulator(bodies=[body], engine=SemiImplicitEulerWithContacts(), dtime=dtime,)
# Initialize the renderer.
renderer = SoftRenderer(camera_mode="look_at", device=device)
camera_distance = 10.0
elevation = 30.0
azimuth = 0.0
renderer.set_eye_from_angles(camera_distance, elevation, azimuth)
# Run the simulation.
writer = imageio.get_writer(outfile, mode="I")
for i in trange(sim_steps):
sim.step()
# print("Body is at:", body.position)
if i % render_every == 0:
rgba = renderer.forward(
body.get_world_vertices().unsqueeze(0), faces, textures
)
img = rgba[0].permute(1, 2, 0).detach().cpu().numpy()
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
"--iters",
type=int,
default=2000,
help="Number of iterations to run optimization for.",
)
parser.add_argument(
"--no-viz", action="store_true", help="Skip visualization steps."
)
args = parser.parse_args()
# Initialize the soft rasterizer.
renderer = SoftRenderer(camera_mode="look_at", device="cuda:0")
# Camera settings.
camera_distance = (
2.0 # Distance of the camera from the origin (i.e., center of the object)
)
elevation = 30.0 # Angle of elevation
azimuth = 0.0 # Azimuth angle
# Directory in which sample data is located.
DATA_DIR = Path(__file__).parent / "sampledata"
# Read in the input mesh. TODO: Add filepath as argument.
mesh = TriangleMesh.from_obj(DATA_DIR / "sphere.obj")
# Output filename to write out a rendered .gif to, showing the progress of optimization.
# Seed RNG for repeatability
torch.manual_seed(args.seed)
device = "cuda:0"
# # Hyperparams
# dtime = 0.1 # 1 / 30
# simsteps = 50
# seed = 123
# numepochs = 100
# compare_every = 10
# log = True
# logdir = Path("cache/bounce2d")
# Initialize the differentiable renderer.
renderer = SoftRenderer(camera_mode="look_at", device=device)
camera_distance = 15.0
elevation = 0.0
azimuth = 0.0
renderer.set_eye_from_angles(camera_distance, elevation, azimuth)
# GT parameters
position_initial_gt = torch.tensor([-7.5, 0.0], device=device)
radius_gt = 0.75
speed_gt = 1.0
height_gt = 5.0
gravity_gt = -9.0
ball2d_gt = BouncingBall2D(
pos=position_initial_gt,
radius=radius_gt,
