turi::object_detection::CheckpointMetadata Struct Reference

#include <toolkits/object_detection/od_model_trainer.hpp>

Public Attributes
size_t	num_predictions = 0
std::string	model_type = ""
float	evaluate_confidence = 0.f
float	predict_confidence = 0.f
float	nms_threshold = 0.f
bool	use_most_confident_class = false

Detailed Description

Stores additional data for specific model backend for a checkpoint.

Definition at line 134 of file od_model_trainer.hpp.

Member Data Documentation

evaluate_confidence

float turi::object_detection::CheckpointMetadata::evaluate_confidence = 0.f

The confidence threshold for evaluation

Definition at line 142 of file od_model_trainer.hpp.

model_type

std::string turi::object_detection::CheckpointMetadata::model_type = ""

The model type name for use in exported models.

Definition at line 139 of file od_model_trainer.hpp.

nms_threshold

float turi::object_detection::CheckpointMetadata::nms_threshold = 0.f

The Non Maximal Suppression threshold for evaluation

Definition at line 148 of file od_model_trainer.hpp.

num_predictions

size_t turi::object_detection::CheckpointMetadata::num_predictions = 0

The number of predictions for the loaded model.

Definition at line 136 of file od_model_trainer.hpp.

predict_confidence

float turi::object_detection::CheckpointMetadata::predict_confidence = 0.f

The confidence threshold for prediction

Definition at line 145 of file od_model_trainer.hpp.

use_most_confident_class

bool turi::object_detection::CheckpointMetadata::use_most_confident_class = false

When true, use NMS only on the most confident class otherwise across all classes.

Definition at line 151 of file od_model_trainer.hpp.

---

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

• toolkits/object_detection/od_model_trainer.hpp