GlammSample<T>

Models & Types · AiDotNet.VisionLanguage.Grounding

A GLaMM training sample: an image, a conversation about it, the regions its <bbox> tokens name, and a mask for every [SEG] in the answers.

Implements: IEquatable<GlammSample<T>>

Constructors

Constructor Summary
GlammSample(Tensor<T> Image, IReadOnlyList<GlammTurn> Turns, IReadOnlyList<double[]> Regions, IReadOnlyList<Tensor<T>> Masks) A GLaMM training sample: an image, a conversation about it, the regions its <bbox> tokens name, and a mask for every [SEG] in the answers.

Properties

Property Summary
IReadOnlyList<GlammTurn> Turns { get; set; } The conversation.
IReadOnlyList<Tensor<T>> Masks { get; set; } Per [SEG] in the answers, in order, its mask [H, W] (1 inside, 0 outside), or null.
IReadOnlyList<double[]> Regions { get; set; } Per <bbox>, in order, its box (x1, y1, x2, y2) as fractions of the image, or null.
Tensor<T> Image { get; set; } The RGB image [3, H, W] with values in 0..1.

Methods

Method Summary
GlammSample<T> <Clone>$()
bool Equals(GlammSample<T> other)
bool Equals(object obj)
bool op_Equality(GlammSample<T> left, GlammSample<T> right)
bool op_Inequality(GlammSample<T> left, GlammSample<T> right)
int GetHashCode()
string ToString()
void Deconstruct(Tensor<T> Image, IReadOnlyList<GlammTurn> Turns, IReadOnlyList<double[]> Regions, IReadOnlyList<Tensor<T>> Masks)