PointNetOptions
Options & Configuration · AiDotNet.Models.Options
Configuration options for PointNet models.
Inheritance: ModelOptions → PointNetOptions
Constructors
| Constructor | Summary |
|---|---|
PointNetOptions() |
Properties
| Property | Summary |
|---|---|
bool UseDropout { get; set; } |
Whether dropout is applied during training. |
bool UseFeatureTransform { get; set; } |
Whether to align per-point features with a learned feature-transform network. |
bool UseInputTransform { get; set; } |
Whether to align the input points with a learned input-transform network. |
double DropoutRate { get; set; } |
Dropout probability applied during training. |
double LearningRate { get; set; } |
Learning rate of the optimizer. |
int InputFeatureDim { get; set; } |
Number of features per input point (3 for xyz coordinates). |
int InputTransformDim { get; set; } |
Dimension of the input transform’s square matrix (3 for xyz). |
int NumClasses { get; set; } |
Number of output classes. |
int[] ClassifierChannels { get; set; } |
Hidden widths of the fully connected classifier head. |
int[] FeatureMlpChannels { get; set; } |
Widths of the shared MLP applied after the feature transform, before global max pooling. |
int[] FeatureTransformFcChannels { get; set; } |
Fully connected widths of the feature-transform network (the 64x64 T-Net). |
int[] FeatureTransformMlpChannels { get; set; } |
Shared-MLP widths of the feature-transform network. |
int[] InputMlpChannels { get; set; } |
Widths of the shared MLP applied after the input transform. |
int[] InputTransformFcChannels { get; set; } |
Fully connected widths of the input-transform network (the 3x3 T-Net). |
int[] InputTransformMlpChannels { get; set; } |
Shared-MLP widths of the input-transform network. |