24th December 2019
Google’s Vision API tool could influence the images we are using on websites
A tool has been created by Google which analyses images using AI and Machine Learning and can tell you what it thinks the image contains or is relevant for.
This tool is part of Google’s Cloud Vision products and is a great indicator of how Google’s AI and Machine Learning algorithms understand images which could help marketers and SEOs to better understand how Google is viewing images we use on websites.
Many tools and search commands that Google provides have not necessarily given an indication of the Google algorithm, so this tool is an exciting development in that regard as it could influence the images that we are placing on websites.
What is Google’s Vision API?
The new image tool is a demo of Google’s Cloud Vision API, a service that allows you to add image analysis to apps and websites.
However, the tool which you can access to test this allows you to upload an image and find out how Google and its machine learning algorithm might interpret the image and what it sees. It analyses the tool and classifies uploaded images, based on the categories below.
The “faces” tab is meant to provide an analysis of the emotion expressed by an image. The accuracy of this result is fairly good, as you can see below in the image of me and Olympic Bobsleigh Pilot Bradley Hall. Although this can get more challenging if more faces are added to the mix and there are a combination of emotions going on within an image.
The object tab shows what objects feature within the image, such as a person, glasses, television, flower, etc. How well this works does depend on the quality of a photo and what objects are in focus.
Looking at the picture we uploaded again, it was able to identify people, but didn’t pick up on the pool table or other items in the background. However, these items are less in focus and not fully visible, so the results from this might be a bit skewed.
The labels tab shows details about the image that the algorithm/Google recognises, which includes features such as ears and mouth but also conceptual aspects like portrait and photography.
This tab highlights words that are associated with that image via the web and can give you an idea of what the web presence of an image says about it. The words and name that come up give an overview of what an image is already associated with, so if you use that image as well Google may already have a preconceived idea of what sort of words and topics that image is connected to.
The properties tab shows which colours are present within the image.
Safe search shows how the image ranks in terms of how unsafe it is and why it might be classified as being unsafe/explicit. The descriptions of potentially unsafe images are as follows:
It will give you an indication of how unlikely/likely it is that an image may be classified under these categories.
What does this mean for SEOs?
The Vision API tool can identify a whole range of things, from a logo to a celebrity’s face. Its ability to recognise logos means it appears to have the ability to recognise text through Optical Character recognition (OCR) which allows it better understand images and use this for ranking purposes.
This means that Google can read the captions, alt text, file name, etc within an image which many of us know do help when focusing on on-page SEO.
Another thing to note is that YouTube thumbnails sometimes rank in Google image search, so you should be running these through the tool too as this can help to ensure they are optimised for the kind of keywords and content which the video is targeting or focusing on.
Overall, whilst some elements of the tool don’t always seem perfect, it is definitely a useful tool for getting an idea of how Google might understand the images which you are using for yourself or for clients. It can also give SEOs as to whether or not they might need to optimise an image better.
If you have multiple potential images which you could use within a page, try running them through this tool to see which Google sees as being most appropriate for your target keywords for that page.
Alternatively, for further advice on how to optimise your website to perform better, get in touch with the experts at Varn today.