Turi Create
4.0
|
#include <toolkits/object_detection/od_darknet_yolo_model_trainer.hpp>
Public Member Functions | |
DarknetYOLOModelTrainer (const DarknetYOLOCheckpoint &checkpoint, neural_net::compute_context *context) | |
std::shared_ptr< neural_net::Publisher< TrainingOutputBatch > > | AsTrainingBatchPublisher (std::unique_ptr< data_iterator > training_data, size_t batch_size, int offset) override |
std::shared_ptr< neural_net::Publisher< EncodedBatch > > | AsInferenceBatchPublisher (std::unique_ptr< data_iterator > test_data, size_t batch_size, float confidence_threshold, float iou_threshold) override |
InferenceOutputBatch | DecodeOutputBatch (EncodedBatch batch, float confidence_threshold, float iou_threshold) override |
std::shared_ptr< neural_net::Publisher< std::unique_ptr< Checkpoint > > > | AsCheckpointPublisher () override |
Subclass of ModelTrainer encapsulating the darknet-yolo architecture.
Definition at line 166 of file od_darknet_yolo_model_trainer.hpp.
turi::object_detection::DarknetYOLOModelTrainer::DarknetYOLOModelTrainer | ( | const DarknetYOLOCheckpoint & | checkpoint, |
neural_net::compute_context * | context | ||
) |
Initializes a model from a checkpoint.
|
overridevirtual |
Returns a publisher that can be used to request checkpoints.
Implements turi::object_detection::ModelTrainer.
|
overridevirtual |
Given a data iterator, return a publisher of inference model outputs.
Implements turi::object_detection::ModelTrainer.
|
overridevirtual |
Given a data iterator, return a publisher of model outputs.
Reimplemented from turi::object_detection::ModelTrainer.
|
overridevirtual |
Convert the raw output of the inference batch publisher into structured predictions.
Implements turi::object_detection::ModelTrainer.