Merge the patched input into a image with sliding window.

Discard class token and reshape 1D feature map to a 2D grid.

Create and load a backbone model given a config. Args: ---- preset: A backbone preset to load pre-defind configs. Returns: -

Inverts a depth map with numerical stability. Args: ---- depth (np.ndarray): Depth map to be inverted. eps (float): Minimum v

Apply Non-Maximum Suppression (NMS) horizontally on the given ratio matrix. Args: ---- ratio (np.ndarray): Input ratio matrix.

Split the input into small patches with sliding window.

Initialize ResidualBlock.

Create a residual block.

Find foreground-background relations between neighboring pixels. Args: ---- d (np.ndarray): Depth matrix. t (float): Threshol

Generate thresholds and weights for the given range. Args: ---- t_min (float): Minimum threshold. t_max (float): Maximum thre

Apply Non-Maximum Suppression (NMS) vertically on the given ratio matrix. Args: ---- ratio (np.ndarray): Input ratio matrix.

Create a 3-level image pyramid.

Calculate Boundary F1 score. Args: ---- pr (np.ndarray): Predicted depth matrix. gt (np.ndarray): Ground truth depth matrix.

Find connected components in the given row and column indices. Args: ---- r (np.ndarray): Row indices. c (np.ndarray): Column

Create a DepthPro model and load weights from `config.checkpoint_uri`. Args: ---- config: The configuration for the DPT model archite

Create and load a VIT backbone module. Args: ---- preset: The VIT preset to load the pre-defined config. use_pretrained: Load

Calculate edge recall for image matting. Args: ---- pr (np.ndarray): Predicted depth matrix. gt (np.ndarray): Ground truth bi

Return exif information as a dictionary. Args: ---- img_pil: A Pillow image. Returns: ------- A dictionary with extr

Find foreground-background relations between neighboring pixels in binary masks. Args: ---- d (np.ndarray): Binary depth matrix.

Find foreground-background relations between neighboring pixels with Non-Maximum Suppression. Args: ---- d (np.ndarray): Depth matrix

Decode by projection and fusion of multi-resolution encodings. Args: ---- x (torch.Tensor): Input image. Returns

Convert a focal length given in mm (35mm film equivalent) to pixels.

Get the Torch device.

Infer depth and fov for a given image. If the image is not at network resolution, it is resized to 1536x1536 and the estimated depth

Load an RGB image. Args: ---- path: The url to the image to load. auto_rotate: Rotate the image based on the EXIF data, defau

Run DepthPro inference example.

Make a ViTb16 backbone for the DPT model.

Resample the ViT patch size to the given one.

Resample the ViT module to the given size.

Run Depth Pro on a sample image.

Calculate Scale-Invariant Boundary F1 Score for depth-based ground-truth. Args: ---- predicted_depth (np.ndarray): Predicted depth ma

Calculate Scale-Invariant Boundary Recall Score for mask-based ground-truth. Args: ---- predicted_depth (np.ndarray): Predicted depth

Initialize DepthPro. Args: ---- encoder: The DepthProEncoder backbone. decoder: The MultiresConvDecoder decod

Initialize multiresolution convolutional decoder. Args: ---- dims_encoder: Expected dims at each level from the encoder.

Initialize feature fusion block. Args: ---- num_features: Input and output dimensions. deconv: Whether to use

Initialize the Field of View estimation block. Args: ---- num_features: Number of features used. fov_encoder:

Initialize DepthProEncoder. The framework 1. creates an image pyramid, 2. generates overlapping patches with a slidin

Decode the multi-resolution encodings.

Apply residual block.

Process and fuse input features.

Forward the fov network. Args: ---- x (torch.Tensor): Input image. lowres_feature (torch.Tensor): Low resolut

Encode input at multiple resolutions. Args: ---- x (torch.Tensor): Input image. Returns: -------

Encode features.

Return the internal image size of the network.

Return the full image size of the SPN network.