Generator that yields (RGB frame, original_shape_tuples, native_fps). If fix_width is provided, resizes the frame preserving aspect ratio before y

Takes full tensors and returns a random temporal slice based on whether the logic is bidirectional. Parameters ----

Convert JSON configuration file to argparse.Namespace object. Parameters ---------- json_path : str Path to the JSON configuratio

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

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

Read .flo file in Middlebury format

Resize flow_gt and mask to target resolution for loss computation.

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

Processes specialized tokens with shape (1, 2, X, C) for multi-frame processing: 1) Uses the first position (index=0) for the first frame onl

Try to move the input variable `t` to a cuda device. Args: t: Input. Returns: t_cuda: `t` moved to a cuda device, if su

Validation step for chairs, sintel, and spring datasets. Parameters ---------- data_blob : tuple[torch.Tensor, torch.Tensor,

Validation step for KITTI dataset. Parameters ---------- data_blob : Tuple[torch.Tensor, torch.Tensor, torch.Tensor]

Generates spatial positions for a batch of patches. Args: batch_size: Number of samples in the batch. height: Height

Init. Args: features (int): number of features

Init. Args: features (int): number of features

Initializes the 2D RoPE module.

Args: img_size (int, tuple): input image size patch_size (int, tuple): patch size in_chans (int): number

Collate function for video tracks data. Returns a batch dictionary and a list of flags.

Configure optimizers and learning rate schedulers. Returns ------- Dict[str, Any] Dictionary containing optimizer

Configure optimizers and learning rate schedulers. Returns ------- Dict[str, Any] Dictionary containing optimizer

Move all contents of a dataclass to cuda inplace if supported. Args: batch: Input dataclass. Returns: batch_cuda: `batc

Used for tensorboard visualization