Google Cloud Platform (GCP) AutoML Vision Product Overview
Google Cloud Platform (GCP)
Codeless model building, built with Neural Architecture Search (NAS) & using NASNets
Usage Pattern (versus Vision API)
Objective - enable developers with no ML expertise
Primary use case - classification
Data requirements - images with labelled data
Output format - labels with probability
Customisation - none
Efforts - low for solution designing
Features
Evaluation metrics
Precision (quality - percent classified correctly) - number correctly classified with a specific label / total number with that label
Recall (quantity - getting a high number of correctly classified instances) - number classified with a specific label / total number with that label
Confusion Matrix - drill down showing images correctly and incorrectly classified
Average Precision (Area Under Precision/Recall Curve)
Train on your own image set