Struct PositionalEncoding 

pub struct PositionalEncoding {
    pub n_freqs: usize,
    pub include_input: bool,
}

Fourier positional encoding used by NeRF.

For a scalar x and n_freqs frequencies the encoding is: [sin(2^0 x), cos(2^0 x), sin(2^1 x), cos(2^1 x), …, sin(2^{L-1} x), cos(2^{L-1} x)] giving a vector of length 2 * n_freqs.

Fields

n_freqs: usize

Number of frequency bands L.

include_input: bool

Whether to include the raw input (identity) in the encoding.

Implementations

impl PositionalEncoding

pub fn new(n_freqs: usize) -> Self

Create a new encoding with n_freqs frequency bands.

pub fn with_identity(n_freqs: usize) -> Self

Create a new encoding including the raw input.

pub fn encode_scalar(&self, x: f64) -> Vec<f64>

Encode a scalar x. Returns a vector of length 2 * n_freqs (or 2 * n_freqs + 1 when include_input is true).

pub fn encode(&self, x: &[f64]) -> Vec<f64>

Encode an arbitrary-length input vector by encoding each element independently and concatenating the results.

pub fn output_dim_scalar(&self) -> usize

Output dimension for a single scalar input.

pub fn output_dim(&self, input_len: usize) -> usize

Output dimension for an input of length input_len.

Trait Implementations

impl Clone for PositionalEncoding

fn clone(&self) -> PositionalEncoding

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for PositionalEncoding

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.