Turi Create  4.0
Public Member Functions | List of all members
turi::object_detection::Checkpoint Class Referenceabstract

#include <toolkits/object_detection/od_model_trainer.hpp>

Public Member Functions

virtual std::unique_ptr< ModelTrainer > CreateModelTrainer (neural_net::compute_context *context) const =0
 
virtual neural_net::pipeline_spec ExportToCoreML (const std::string &input_name, const std::string &coordinates_name, const std::string &confidence_name, bool use_nms_layer, float iou_threshold, float confidence_threshold) const =0
 

Detailed Description

A representation of all the parameters needed to reconstruct a model.

Definition at line 159 of file od_model_trainer.hpp.

Member Function Documentation

◆ CreateModelTrainer()

virtual std::unique_ptr<ModelTrainer> turi::object_detection::Checkpoint::CreateModelTrainer ( neural_net::compute_context *  context) const
pure virtual

Loads the checkpoint into an active ModelTrainer instance.

Implemented in turi::object_detection::DarknetYOLOCheckpoint.

◆ ExportToCoreML()

virtual neural_net::pipeline_spec turi::object_detection::Checkpoint::ExportToCoreML ( const std::string &  input_name,
const std::string &  coordinates_name,
const std::string &  confidence_name,
bool  use_nms_layer,
float  iou_threshold,
float  confidence_threshold 
) const
pure virtual

Returns the CoreML spec corresponding to the current model.

The result must be a pipeline that accepts an image input and yields at least two outputs, all with the given names. The outputs must be suitable for passing directly into a NonMaximumSuppression model.

Implemented in turi::object_detection::DarknetYOLOCheckpoint.

The documentation for this class was generated from the following file: