PICKOptions

Options & Configuration · AiDotNet.Document.Options

Configuration options for the PICK document model (Yu et al., ICPR 2020). The defaults are the reference implementation’s configuration (wenwenyu/PICK-pytorch, config.json).

Inheritance: ModelOptions → NeuralNetworkOptions → ModelHyperparameterOptions → DocumentNeuralNetworkOptions → PICKOptions

Constructors

Constructor Summary
PICKOptions() Initializes a new instance of the PICKOptions class with the reference configuration.

Properties

Property Summary
double GraphEta { get; set; } Gets or sets eta, the weight of node distance in the graph-learning loss.
double GraphGamma { get; set; } Gets or sets gamma, the weight of the adjacency’s Frobenius norm in the graph-learning loss.
double GraphLossWeight { get; set; } Gets or sets the graph-learning loss weight added to the CRF loss (gl_loss_lambda).
int FeedForwardDim { get; set; } Gets or sets the text encoder’s feed-forward width.
int GraphLearningDim { get; set; } Gets or sets the graph-learning projection width.
int ImageFeatureDim { get; set; } Gets or sets the image backbone’s output channels before the ROI head.
int LstmHiddenDim { get; set; } Gets or sets the BiLSTM hidden size per direction.
int LstmLayers { get; set; } Gets or sets the BiLSTM depth.
int NumEntityTypes { get; set; } Gets or sets the number of IOB tags the CRF predicts.
int NumGcnLayers { get; set; } Gets or sets the graph convolution depth.
int[] ImageEncoderDepths { get; set; } Gets or sets the bottleneck blocks per ResNet stage of the image branch.

Methods

Method Summary
void Validate() Throws if a value this model requires has been left unset or is not positive.