Function of ABEJA Platform

Introduction

This page explains the functions of ABEJA Platform. ABEJA Platform provides functions for each task required for machine learning.

DataLake

• Features for storing and retrieving data
• Data source function allows data input / output via API
• Cooperation with annotation tool. Create supervised data efficiently

Annotation

• GUI can be used to selecting templates, set labels, and export to datasets
• Supports repetitive templates such as images, audio, video, and text
• With the review function, work can be shared by roles such as workers and reviewers
• Supports pre-inference function

Dataset

• Manage raw data and annotated data uploaded to the datalake
• You can also save your own annotated data

Training

• You can be collectively managed with GUI (Training job, Parameters, and Accuracy results, etc.)
• Various frameworks can be used
• Jyupyter Notebook
• Local Training features
• Template features

Model / Deploy

• Model version control
• Web API Deployment
• Deploy multiple versions of the API
• Blue/Green Deployment
• Edge Device Deployment

Monitoring / Trigger

• Training log / Web API log
• API monitoring with metrics
• Cooperation with input data by trigger function