def __init__(self, params_dict):
super(VoxHonet, self).__init__(params_dict)
self.voxel_grid_side = params_dict['voxel_grid_side']
self.load_pcls(params_dict['dataset_folder'])
self.num_joints = 16
self.main_loss_conv = cudafy(
VargoNet.VargoNetConvBlock(kernel_size=3,
stride=1,
filters=21,
in_channels=256,
padding=1), self.use_cuda)
self.main_loss_deconv1 = cudafy(
nn.Upsample(size=self.crop_res, mode='bilinear'), self.use_cuda)
if self.cross_entropy:
self.softmax_final = cudafy(VargoNet.SoftmaxLogProbability2D(),
self.use_cuda)
self.innerproduct1_joint1 = cudafy(
nn.Linear(in_features=524288, out_features=200), self.use_cuda)
self.innerproduct2_joint1 = cudafy(
nn.Linear(in_features=200, out_features=(self.num_joints - 1) * 3),
self.use_cuda)
self.innerproduct1_joint2 = cudafy(
nn.Linear(in_features=262144, out_features=200), self.use_cuda)
self.innerproduct2_joint2 = cudafy(
nn.Linear(in_features=200, out_features=(self.num_joints - 1) * 3),
self.use_cuda)
self.innerproduct1_joint3 = cudafy(
nn.Linear(in_features=131072, out_features=200), self.use_cuda)
self.innerproduct2_joint3 = cudafy(
nn.Linear(in_features=200, out_features=(self.num_joints - 1) * 3),
self.use_cuda)
self.innerproduct1_joint_main = cudafy(
nn.Linear(in_features=65536, out_features=200), self.use_cuda)
self.innerproduct2_join_main = cudafy(
nn.Linear(in_features=200, out_features=200), self.use_cuda)
self.obj_voxel_in = cudafy(
nn.Conv3d(in_channels=1, out_channels=1, kernel_size=6, stride=2),
self.use_cuda)
self.obj_conv3d0 = cudafy(
nn.Conv3d(in_channels=1, out_channels=1, kernel_size=5, stride=2),
self.use_cuda)
self.obj_conv3d1 = cudafy(
nn.Conv3d(in_channels=1, out_channels=1, kernel_size=5),
self.use_cuda)
self.funnel_in0 = cudafy(
nn.Linear(in_features=200 + 216, out_features=100), self.use_cuda)
self.funnel_in1 = cudafy(nn.Linear(in_features=100, out_features=100),
self.use_cuda)
self.funnel_in2 = cudafy(
nn.Linear(in_features=100, out_features=(self.num_joints - 1) * 3),
self.use_cuda)
def __init__(self, params_dict):
super(HONet, self).__init__(params_dict)
self.num_joints = 16
self.main_loss_conv = cudafy(
VargoNet.VargoNetConvBlock(kernel_size=3,
stride=1,
filters=21,
in_channels=256,
padding=1), self.use_cuda)
self.main_loss_deconv1 = cudafy(
nn.Upsample(size=self.crop_res, mode='bilinear'), self.use_cuda)
if self.cross_entropy:
self.softmax_final = cudafy(VargoNet.SoftmaxLogProbability2D(),
self.use_cuda)
self.innerproduct1_joint1 = cudafy(
nn.Linear(in_features=524288, out_features=200), self.use_cuda)
self.innerproduct2_joint1 = cudafy(
nn.Linear(in_features=200, out_features=(self.num_joints - 1) * 3),
self.use_cuda)
self.innerproduct1_joint2 = cudafy(
nn.Linear(in_features=262144, out_features=200), self.use_cuda)
self.innerproduct2_joint2 = cudafy(
nn.Linear(in_features=200, out_features=(self.num_joints - 1) * 3),
self.use_cuda)
self.innerproduct1_joint3 = cudafy(
nn.Linear(in_features=131072, out_features=200), self.use_cuda)
self.innerproduct2_joint3 = cudafy(
nn.Linear(in_features=200, out_features=(self.num_joints - 1) * 3),
self.use_cuda)
self.innerproduct1_joint_main = cudafy(
nn.Linear(in_features=65536, out_features=200), self.use_cuda)
self.innerproduct2_join_main = cudafy(
nn.Linear(in_features=200, out_features=200), self.use_cuda)
self.obj0 = cudafy(
nn.Linear(in_features=self.size_obj_input,
out_features=self.size_obj_input), self.use_cuda)
self.obj1 = cudafy(
nn.Linear(in_features=self.size_obj_input,
out_features=self.size_obj_input), self.use_cuda)
self.obj2 = cudafy(
nn.Linear(in_features=self.size_obj_input,
out_features=self.size_obj_input), self.use_cuda)
self.merge_hand_obj = cudafy(
nn.Linear(in_features=200 + self.size_obj_input,
out_features=200 + self.size_obj_input), self.use_cuda)
self.hand_obj0 = cudafy(
nn.Linear(in_features=200 + self.size_obj_input, out_features=100),
self.use_cuda)
self.hand_obj1 = cudafy(
nn.Linear(in_features=100, out_features=(self.num_joints - 1) * 3),
self.use_cuda)
def map_out_conv(self, in_channels):
return cudafy(
VargoNet.VargoNetConvBlock(kernel_size=3,
stride=1,
filters=21,
in_channels=in_channels,
padding=1), self.use_cuda)