OCI - Vision

The very common use case seen today is image recognition and detecting objects within that image. The image could have document with texts or it could have any objects. Its important to not only detect the object and also perform analysis of the object. For example if it is an objects then its important to classify the object, if it is a document then classify it whether it is invoice or receipt and what language it is written.

OCI Vision is an AI powered service and it performs all of the above. It split between document AI for document-centric image and perform image analysis using machine learning capability for the objects. We will see how it performs in this article.

Prerequisites:

OCI account, it could be trail or paid account.
Account credentials to set the policy. The user should have privilege to use ai-service-vision-family.
Objects to analyse, it could be in the local storage or OCI object storage. If the location is OCI Object storage, then necessary privilege is granted to access the object storage files.

Login into OCI console.

Vision service is located under Analytics & AI -> AI services -> Vision.

Once we choose the Vision, it will open up Vision overview page.

We have five options to perform with this service.
Image classification:

It takes the image from Local storage or object storage and then analyse and classifies it into different types based on visual appearance, objects identified in the image. It also provides the confidence level through scores for each identified object. We could assume that object with high score have been identified correctly.

Lets discuss with an example.

Chosen an random image located in local storage.

The Results shows it has identified Dice in the image and its confidence score is 99.97% and it classifies the object into game, sports and also its Board game. Its not completely blue colour, hence the score for the colour is less.

Object Detection:

This feature detects all the objects located in an image. It provides confidence score for each object it has identified.

In the results we could see the objects identified. Similar objects are identified with different score. The object detection feature has identified only the objects in the image.

 

Face Detection:

This feature detects how many human faces detected in a picture.

The object here is an image with cat. We could the Empty results returned.

The image used in the object detection has been used here. We could see Face detection has said Empty results returned.

Text Detection:

 The image used here has text and face. Text analysis will identify the text in the image. The results will be categorized as Line and Words.

 

Video Analysis:

 Its one of the very useful use case. This feature will analyze the video and reports about the Labels, Objects, Text and Face.