Hi~, your work is awesome! but I have a question about that GS_Refiner how to optimize the _delta_R and _delta_T with not inputting the renderer or adding to the init_camera? Is there some magic in the code?
`
def multiple_refine_pose_with_GS_refiner(obj_data, init_pose, gaussians, device):
def GS_Refiner(image, mask, init_camera, gaussians, return_loss=False):
if image.dim() == 4:
image = image.squeeze(0)
if image.shape[2] == 3:
image = image.permute(2, 0, 1) # 3xSxS
if mask is None:
mask = torch.ones_like(image[0])
if mask.dim() == 2:
mask = mask[None, :, :]
if mask.dim() == 4:
mask = mask.squeeze(0)
if mask.shape[2] == 1:
mask = mask.permute(2, 0, 1) # 1xSxS
Hi~, your work is awesome! but I have a question about that GS_Refiner how to optimize the _delta_R and _delta_T with not inputting the renderer or adding to the init_camera? Is there some magic in the code?
` def multiple_refine_pose_with_GS_refiner(obj_data, init_pose, gaussians, device): def GS_Refiner(image, mask, init_camera, gaussians, return_loss=False): if image.dim() == 4: image = image.squeeze(0) if image.shape[2] == 3: image = image.permute(2, 0, 1) # 3xSxS if mask is None: mask = torch.ones_like(image[0]) if mask.dim() == 2: mask = mask[None, :, :] if mask.dim() == 4: mask = mask.squeeze(0) if mask.shape[2] == 1: mask = mask.permute(2, 0, 1) # 1xSxS