Create label tensors with the same size as the input. Parameters: prediction (tensor) - - tpyically the prediction from a discrim

Inference interface for the model Args: x (torch.Tensor): input tensor of shape (b, c, h, w) pad_input (bool,

Inference interface for the model for PIL image Args: pil_img (PIL.Image.Image): input PIL image pad_input (b

Inference interface for the model with horizontal flip augmentation Horizontal flip augmentation improves the accuracy of the model b

Initialize network weights. Parameters: net (network) -- network to be initialized init_type (str) -- the name of an initializa

Define the common options that are used in both training and test.

To calculate the distance between each two depth maps.

Ensure that the depth model is loaded

log(nCk) using stirling approximation

Detects if DepthMap was installed as a stable-diffusion-webui script, but run without current directory set to the stable-diffusion-webui root. Ch

Move to RAM to conserve VRAM

Calculate losses, gradients, and update network weights; called in every training iteration

360 depth to world points given 2D depth is an equirectangular projection of a spherical image Treat depth as radius longitude : -pi

Parse our options, create checkpoints directory suffix, and set up gpu device.

Parse a list of values for the key if the value is a string. The values are separated by a comma. Modifies the config in place.

Generate positional encodings Args: sequence_length (int): Sequence length embedding_dim (int): Embedding dimension

print current losses on console; also save the losses to the disk Parameters: epoch (int) -- current epoch iters (int

Print the total number of parameters in the network and (if verbose) network architecture Parameters: verbose (bool) -- if verbos

Reconstruct depth to 3D pointcloud with the provided focal length. Return: pcd: N X 3 array, point cloud

Undoes offload

cycles through iterables but sample wise first yield first sample from first iterable then first sample from second iterable and so on th

Reset the self.saved status

Resize numpy (or image read by imageio) depth map Args: depth (numpy): depth width (int): image width height (int): image

Resize image to limit maximum edge length while keeping aspect ratio Args: img (Image.Image): Image to be resized max_edge_r

Constructs a ResNet-152 model. Args: pretrained (bool): If True, returns a model pre-trained on ImageNet

Save an RGB point cloud as a PLY file. :paras @pcd: Nx3 matrix, the XYZ coordinates @rgb: NX3 matrix, the rgb colors for each 3D poin

Scale the image and output it in torch.tensor. :param img: input rgb is in shape [H, W, C], input depth/disp is in shape [H, W] :param sc

AdaHessian Optimizer Lifted from https://github.com/BIGBALLON/distribuuuu/blob/master/distribuuuu/utils.py Originally licensed MIT, Copyright

Perform an individual depth prediction without ensembling. Args: rgb_in (torch.Tensor): Input RGB image.

Splits the arguments that are combined with '__' into multiple arguments. Combined arguments should have equal number of keys and values.

Modification of https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/levit.py: stem_b16 such that ConvTranspose2