Struct VideoReader 

pub struct VideoReader<B: Backend> { /* private fields */ }

A high-level video reader that loads frames into memory.

Supports animated images (GIF, APNG), image sequences, and basic container formats. All frames are cached for fast random access.

Implementations

impl<B: Backend> VideoReader<B>

pub fn open(path: impl AsRef<Path>, options: &VideoOpenOptions) -> Result<Self>

Opens a video file and loads all frames into memory.

pub fn from_frames(frames: Vec<Frame<B>>, metadata: VideoMetadata) -> Self

Creates a reader from pre-loaded frames and metadata.

Useful for in-memory video processing pipelines where frames are generated or loaded programmatically.

pub fn frames(&self) -> &[Frame<B>]

Returns a reference to all loaded frames.

pub fn frame_count(&self) -> usize

Returns the total number of loaded frames.

pub fn metadata(&self) -> &VideoMetadata

Returns video metadata.

pub fn source_path(&self) -> Option<&Path>

Returns the source file path.

pub fn iter(&self) -> FrameIterator<B>

Returns a frame iterator for sequential access.

pub fn loop_iter(&self) -> FrameIterator<B>

Returns a looping frame iterator.

pub fn get_frame(&self, index: usize) -> Result<&Frame<B>>

Gets a specific frame by index.

pub fn get_range(&self, start: usize, end: usize) -> Result<&[Frame<B>]>

Gets frames in a range.

pub fn seek_to_time(&self, time: Duration) -> Result<&Frame<B>>

Seeks to a specific timestamp and returns the frame.

pub fn to_batch_tensor(&self) -> Result<Tensor<B, 4>>

Converts frames to a 4D tensor [B, C, H, W] (batch dimension).

pub fn frame_differences(&self) -> Result<Vec<Tensor<B, 3>>>

Computes per-frame differences (absolute difference between consecutive frames).

pub fn motion_magnitudes(&self) -> Result<Vec<f32>>

Computes optical flow magnitude (sum of absolute differences) between consecutive frames.