Occlusion augmentation

Applies the flow color wheel to (possibly clipped) flow components u and v. According to the C++ source code of Daniel Scharstein Accord

Estimate optical flow between pair of frames

Get all model parameters recursively.

Initializes an empty video with expected settings.

Flow is represented as difference between two coordinate grids flow = coords1 - coords0

Lists all images from the given directory. NOTE: Do NOT support finding images recursively. Args: directory: The directory to find i

Loads an image from disk. NOTE: This function will always return an image with `RGB` channel order for color image and pixel range [0, 255].

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

Parses an object to a pair of image size, i.e., (height, width). Args: obj: The input object to parse image size from. Returns:

Read .flo file in Middlebury format

Resizes image. This is a wrap of `cv2.resize()`. NOTE: The channel order of the input image will not be changed. Args: image: I

Sets the video codec.

Sets the CRF (constant rate factor) of the video.

Sets the FPS (frame per second) of the video.

Sets the video pixel format.

Upsample flow field [H/8, W/8, 2] -> [H, W, 2] using convex combination

Visualizes a list of image files.

Initializes the video visualizer. Args: path: Path to write the video. (default: None) frame_size: Frame size, i.e.,

D: number of layers for density (sigma) encoder W: number of hidden units in each layer in_channels_xyz: number of input chan

Defines a function that embeds x to (x, sin(2^k x), cos(2^k x), ...) in_channels: number of input channels (3 for both xyz and direct

Defines a function that embeds x to (x, sin(2^k x), cos(2^k x), ...) in_channels: number of input channels (3 for both xyz and direct

D: number of layers for density (sigma) encoder W: number of hidden units in each layer in_channels_xyz: number of input chan