YOLOv9Backbone<T>
Models & Types · AiDotNet.ComputerVision.Detection.ObjectDetection.YOLO
YOLOv9’s backbone (Wang et al. 2024): GELAN stages from the size’s yaml, plus the PGI auxiliary branch that supervises it during training. GELAN-e’s main path is itself a composite (dual) backbone.
Inheritance: NeuralNetworkBase<T> → YOLOv9Backbone<T>
Implements: IDetectionBackbone<T>
Constructors
| Constructor | Summary |
|---|---|
YOLOv9Backbone(YoloBackboneOptions options) |
Creates the backbone for a model size (Nano = t, Small = s, Medium = m, Large = c, XLarge = e). |
Properties
| Property | Summary |
|---|---|
IReadOnlyList<int> AuxiliaryChannels { get; } |
Channels of the auxiliary A3, A4 and A5 maps. |
IReadOnlyList<int> OutputChannels { get; } |
Channels of P3, P4 and P5 fed to the neck. |
IReadOnlyList<int> Strides { get; } |
Strides of P3, P4 and P5. |
bool IsFrozen { get; set; } |
Whether the backbone is frozen. |
string Name { get; } |
The backbone name. |
Methods
| Method | Summary |
|---|---|
IFullModel<T, Tensor<T>, Tensor<T>> WithParameters(Vector<T> parameters) |
|
IReadOnlyList<Tensor<T>> GetFeatureMaps(Tensor<T> input) |
|
List<Tensor<T>> ExtractFeatures(Tensor<T> input) |
|
ModelMetadata<T> GetModelMetadata() |
|
ValueTuple<int, int> GetExpectedInputSize() |
YOLOv9’s canonical training resolution. |
void Freeze() |
Freezes the backbone. |
void ReadParameters(BinaryReader reader) |
|
void Unfreeze() |
Unfreezes the backbone. |
void WriteParameters(BinaryWriter writer) |