Template Deployment (Object Detection)


This page explains how to create an inference API for Object Detection without coding using ABEJA Template.

Prepare a learning model

Prepare a learning model for Object Detection. Here’s how to create a Training model using ABEJA Template (Object detection).

Manage models

Click “Create Model” from “Model” in the console to create a model.

In the model creation screen, select the following items.

Item Description
Name The name of the model
Description Model description
Version Model management version
Deploy after createing Check if you want to deploy at the time the model is created
Deployment name When deploying, enter the deployment name
Templates Select 「Object Detection(CPU)」
Job Definition Select the job definition used during learning
Training Job Select the learning job you want to deploy

Creating an HTTP service

When you go to “Deployment” in the console, It is created with the deployment name you created earlier.

Click the name of the deployment and click Create HTTP Service. Select the version, instance type, and number of instances.

After waiting a few minutes, the status becomes “Available” and the model is deployed as a Web API. This completes deployment. During the deployment process, you can check the log at the time of deployment by selecting the “Logs” button.

Operation check

Click “✔︎Confirm” to test. Here, we will check the operation with a sample “cat” photo.

Select the “Cat” image and click “Send” to return the inference result. In this case, you can see that the coordinate information of the boxes and the inference result is returned as 86.93% Scores for “cat” as Classes.

This page introduced the deployment method using ABEJA Template (Object Detection).