CityscapesDataLoader<T>
Models & Types · AiDotNet.Data.Vision.Benchmarks
Loads the Cityscapes semantic-segmentation dataset (Cordts et al. 2016).
Inheritance: DataLoaderBase<T> → InputOutputDataLoaderBase<T, Tensor<T>, Tensor<T>> → CityscapesDataLoader<T>
How It Works
Expects the canonical extracted layout (after manually downloading both archives from cityscapes-dataset.com — sign-up required, no auto-download):
Output shape: features Tensor[N, H, W, 3] (resized RGB image,
normalized to [0, 1]); labels Tensor[N, H, W] (integer class IDs per
pixel — the standard semantic-seg label format used by PyTorch /
TF / SegFormer / DeepLab loss heads). With MapToTrainIds
(default) source IDs 0..33 are remapped to the 19 evaluation classes;
“ignore” pixels get sentinel value 255.
Constructors
| Constructor | Summary |
|---|---|
CityscapesDataLoader(CityscapesDataLoaderOptions options) |
Properties
| Property | Summary |
|---|---|
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) |
Fields
| Field | Summary |
|---|---|
int IgnoreLabel |
Sentinel ID for “ignore” pixels (background, out-of-eval classes). |
int NumEvalClasses |
Number of evaluation classes when MapToTrainIds is true. |