VideoFrameDataset<T>

Models & Types · AiDotNet.Data.Video

Loads videos represented as directories of sequentially numbered image frames.

Inheritance: DataLoaderBase<T> → InputOutputDataLoaderBase<T, Tensor<T>, Tensor<T>> → VideoFrameDataset<T>

How It Works

This loader expects videos stored as frame directories (the standard preprocessing format for ML video datasets like UCF-101, Kinetics, HMDB51, etc.). Each video is a directory containing sequentially numbered image frames (BMP, PPM, or PGM format).

Expected structure:

Frames are sampled uniformly across the video duration to produce a fixed number of frames per video. Output tensor shape is [N, FramesPerVideo, Height, Width, Channels].

Constructors

Constructor Summary
VideoFrameDataset(VideoFrameDatasetOptions options) Creates a new VideoFrameDataset with the specified options.

Properties

Property Summary
IReadOnlyList<string> ClassNames { get; } Gets the class names discovered from directory names.
int FeatureCount { get; }
int OutputDimension { get; }
int TotalCount { get; }
string Description { get; }
string Name { get; }

Methods

Method Summary
ValueTuple<IInputOutputDataLoader<T, Tensor<T>, Tensor<T>>, IInputOutputDataLoader<T, Tensor<T>, Tensor<T>>, IInputOutputDataLoader<T, Tensor<T>, Tensor<T>>> Split(double trainRatio, double validationRatio, int? seed)