📖Reference

Detailed reference

class Babylog

Used to log information about image predictions made by a computer vision model. The class has several features, including the ability to save log information locally, in the cloud (on Amazon S3), the ability to stream log information, and the ability to use multiple worker threads to handle logging tasks.

• The __init__ method is used to initialize the class and set various properties, such as the configuration path, whether to save logs locally and/or in the cloud, and whether to stream log information. The method also creates a thread pool executor, which is used to handle logging tasks.

Argument	Type	Description	Required	Default Value
config_path	str	path to the yaml config file	Required	N/A
save_local	bool	setting to save logs locally	Optional	True
save_cloud	bool	setting to save logs on the cloud	Optional	False
stream	bool	setting to stream logs via tcp	Optional	False

• The log method is used to log information about an image prediction. The method takes a number of arguments, including the image, the model type, the model name and version, and various other information about the prediction, such as the prediction itself, the timestamp, the latency, and the inference device. The method submits a logging task to the thread pool executor, which will handle the task in the background.

Argument	Type	Description	Required	Default Value	Examples
image	np.ndarray	raw image to be logged	Required	N/A	N/A
model_type	VisionModelType	type of vision model	Required	N/A	(VisionModelType.DETECTION, VisionModelType.CLASSIFICATION)
model_name	str	name of model used	Required	N/A	'resnet50_'inetuned
model_version	str	version of model used	Required	N/A	'1.0.0'
home_dir	str	directory for local logging	Optional	'./babylog/'	'./babylog/'
prediction	np.ndarray	prediction image (annotated...)	Optional	None	N/A
timestamp	int	timestamp	Optional	None	1674993193473
latency	int	inference time in ms	Optional	None	10
inference_device	InferenceDevice	type of inference device	Optional	None	(InferenceDevice.CPU, InferenceDevice.GPU, InferenceDevice.CUDA)
classification	Dict[str, float]	dictionary containing classification information	Optional	None	{"dog": 0.8, "cat": 0.2 }
detection	List[Dict]	dictionary containing detection information	Optional	None	[ { "x": 10, "y": 10, "width": 100, "height": 220, "confidence": 0.97, "classification": {"dog": 0.8, "cat": 0.2}, }, { "x": 20, "y": 10, "width": 100, "height": 220, "confidence": 0.97, "classification": {"dog": 0.8, "cat": 0.2}, }, ]
error_message	str	error message while inference was done	Optional	''	"Out of memory error"

class LoggedPrediction

Used to deserialize a logged prediction and return logged information(image, annotation, errors,...)

• the from_pathmethod is a class method to load a saved serialized message('.bin') from filepath of type str

• the from_bytesmethod is a class method to load a serialized message from binary_ of type bytes

• LoggedPrediction:

Property	Type	Description	Examples
model	Dict	dictionary storing model information	{ 'type': 'DETECTION', 'version': '1.0.0', 'name': 'resnet50_finetuned' }
inference_stats	Dict	dictionary for storing inference information	{ 'latency': 50, 'inferenceDevice': 'CPU', 'errorMessage': '' }
device_details	Dict	dictionary for storing device details	{ 'deviceName': 'DEVICE_NAME', 'groupName': 'GROUP_NAME' }
timestamp	int	timestamp	1674993193473
classification	List[Dict]	dictionary containing classification information	[ {'className': 'dog', 'probability': 0.5}, {'className': 'cat', 'probability': 0.5} ]
detection	List[Dict]	dictionary containing detection information	[{'x': 1, 'y': 1, 'width': 100, 'height': 220, 'confidence': 0.97, 'classificationResult': [{'className': 'dog', 'probability': 0.8}, {'className': 'cat', 'probability': 0.2}]}, {'x': 1, 'y': 1, 'width': 50, 'height': 150, 'confidence': 0.97, 'classificationResult': [{'className': 'dog', 'probability': 0.5}, {'className': 'cat', 'probability': 0.5}]}]
image	np.ndarray	logged raw image	N/A
prediction	np.ndarray	logged prediction image	N/A