Generates a color wheel for optical flow visualization as presented in: Baker et al. "A Database and Evaluation Methodology for Optical F

Resize 2D map by nearest interpolation. Return the nearest neighbor index and mask of the resized map. ### Parameters - `inputs`: a sing

Resize 2D map by nearest interpolation. Return the nearest neighbor index and mask of the resized map. ### Parameters - `inputs`: a sing

Convert rotations given as rotation matrices to quaternions. Args: matrix: Rotation matrices as tensor of shape (..., 3, 3). Re

More memory-friendly matching

Generate the mesh grid, centering at zero. Args: kernel_size (int): Returns: xy (ndarray): with the shape (kernel_size, kernel

Modulate the input tensor using scaling and shifting parameters.

Modulate the input tensor using scaling and shifting parameters.

Returns: (H, W) array, a mask for all pixels that have a prediction

normalize the image for depth anything input args: x: the input images [B T C H W]

Args: boxes (ndarray): an Nx5 numpy array of (x_center, y_center, width, height, angle_degrees) format

Calculate PDF of the bivariate Gaussian distribution. Args: sigma_matrix (ndarray): with the shape (2, 2) grid (ndarray): generate

Pool tokens in x using mask. NOTE: We assume x does not require gradients. Args: x: (B, L, D) tensor of tokens.

Cat x with a positional encoding of x with scales 2^[min_deg, max_deg-1]. Instead of computing [sin(x), cos(x)], we use the trig identity cos(

Convert 2D position grid (HxWx2) to sinusoidal embeddings (HxWxC) Args: pos_grid: Tensor of shape (H, W, 2) containing 2D coordinate

Run the model on one video and return per-frame probabilities in [0,1], length T. Matches the repo's behavior by taking center 50 fra

Process the pair of inputs and outputs. If they contain batches, the pairs can be consumed one-by-one using `zip`: .. code-b

Convert a PyCOLMAP Reconstruction Object to batched PyTorch tensors. Args: reconstruction (pycolmap.Reconstruction): The reconstruct

Quaternion Order: XYZW or say ijkr, scalar-last Convert rotations given as quaternions to rotation matrices. Args: quaternions:

Register a "standard" version of ADE20k panoptic segmentation dataset named `name`. The dictionaries in this registered dataset follows detec

Preparation for a new round of evaluation. Should be called before starting a round of evaluation.

r""" Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles custom timesteps. Any kwa

r""" Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles custom timesteps. Any kwa

Forward pass through the fusion blocks. Args: features (List[Tensor]): List of feature maps from different layers.

Loss function defined over sequence of flow predictions

Loss function defined over sequence of flow predictions

Loss for classifying if a point is within pixel threshold of its target.

Calculate the rotated sigma matrix (two dimensional matrix). Args: sig_x (float): sig_y (float): theta (float): Radian mea

Perform bundle adjustment optimization (compatible with pycolmap 0.5+) Args: reconstruction: pycolmap.Reconstruction object

Convert a unit quaternion to a standard form: one in which the real part is non negative. Args: quaternions: Quaternions with re

Convert a unit quaternion to a standard form: one in which the real part is non negative. Args: quaternions: Quaternions with re

r""" Decode a batch of images using a tiled decoder. Args: z (`torch.Tensor`): Input batch of latent vectors.

r"""Encode a batch of images using a tiled encoder. Args: x (`torch.Tensor`): Input batch of videos. Returns:

Drops labels to enable classifier-free guidance.

tracks3d: B T N 3 dyn_prob: B T N 1

Iteratively refine camera pose predictions. Args: pose_tokens (torch.Tensor): Normalized camera tokens with shape [B, 1,