Vision - Detection 2D

Category: task.vision.detection.2d
Version: 1.0.1
UUID: abf9630dac8e160cae221502f4adcfa741301716a667d6aa3e892566ffb166c2
Execution context: leip_af
Choice Priority: 1000.0

Value Parameters

Name	Synonyms	Type	Values	Help
Training Batch Size	train.batch_size_train	scalar	int, min: 1
Validation Batch Size	train.batch_size_val	scalar	int, min: 1
Random Number Generator Seed	global_seed, task.global_seed, experiment.seed	scalar	int
Target Image Height	target.height	scalar	int, min: 1
Experiment Name	experiment.name	scalar	string
Normalization Scheme	data.normalization	choice	imagenet inception yolo	The normalization is applied at augmentation time. These names are proxies to the mean and standard deviation to be applied to the images in the dataset. The yolo option applies no normalization at all.
Number of parallel workers	train.num_workers	scalar	int, min: 0
Target Image Width	target.width	scalar	int

Constraints

1. Data output width has to match model input width (model.input.width == target.width)
2. Data output height has to match model input height (model.input.height == target.height)
3. Model has to be a 2D detection model: model \(\in\) model.detection.2d
4. Data generator has to be a 2D detection data generator: data_generator \(\in\) data_generator.vision.detection.2d
5. Evaluation has use a 2D detection metric: metric \(\in\) metric.detection.2d
6. Export Data has to use a 2D detection format: export_data \(\in\) export_data.detection.2d
7. Data report has to use a 2D detection format: data_report \(\in\) data_report.detection.2d

This component fits into

Name	UUID	Synonyms
Full Recipe	4511dd...	task