VideoFrameDatasetOptions
Options & Configuration · AiDotNet.Data.Video
Configuration options for the VideoFrameDataset.
How It Works
Videos are represented as directories of sequentially numbered image frames. This is the standard format for preprocessed ML video datasets (UCF-101, Kinetics, etc.). Structure: root/class_name/video_name/frame_001.bmp, frame_002.bmp, …
Constructors
| Constructor | Summary |
|---|---|
VideoFrameDatasetOptions() |
Properties
| Property | Summary |
|---|---|
bool NormalizePixels { get; set; } |
Whether to normalize pixel values to [0, 1]. |
bool UseDirectoryLabels { get; set; } |
Whether class labels are determined by parent directory names. |
int Channels { get; set; } |
Number of color channels per frame. |
int FrameHeight { get; set; } |
Target frame height in pixels. |
int FrameWidth { get; set; } |
Target frame width in pixels. |
int FramesPerVideo { get; set; } |
Number of frames to sample per video. |
int? MaxSamples { get; set; } |
Optional maximum number of videos to load. |
int? RandomSeed { get; set; } |
Optional random seed for reproducible sampling. |
string RootDirectory { get; set; } |
Root directory containing class subdirectories, each with video subdirectories containing frame images. |
string[] FrameExtensions { get; set; } |
File extensions for frame images. |