Using AI to Tag your Pictures

Posted by on June 9, 2018

Visual is better.

In today’s online market place, visual marketing is an integral part of reaching out to customers. Research show that the human brain is more visual. Up to 80% of people remember what they see compared to only 20% of what they read.

This is one of the main reasons why visual content resonates more than written content. It gets shared more. It has a greater chance of going viral.

Visual has issues.

But by themselves, digital images have no inherent textual elements or semantic description. This creates challenges for machines.
Machines find it difficult to understand images. Google searches can’t use the visuals to refer clients. Google works by using textual queries and applicable keywords or phrases to find you. Without the textual context, your images fail to help your buyers find you. Potential buyers won’t “see” your product.

So even if you have great pictures of your “neon green sweat shirt” or a “red striped truck”. If they are not matched with text or tags, they are not searchable.

Visual + Tags.

Images are great for going viral. To make it even better, tag it. Make your images searchable online. Tag them with the relevant text or keywords. This helps you improve your search rankings. While your images grab your buyers’ attention, the tags let search engines refer clients. So, you connect with more customers.

Visual+Tags+AI

But is it scalable? If you have thousands of images, it would take a lot of time and money to tag and classify these by hand. Enter Artificial intelligence.

artificial intelligence

artificial intelligence

A good example would be the use of CAPTCHAs. These Captchas use distorted images to tell humans or bots apart. These images are designed to fool automated bots – but not humans. So traditionally, spammers use a lot of humans to defeat Captchas.

AI can be used to defeat Captchas. But AI needs a human curated training set. The training set matches the images with relevant text. AI then creates a predictive model.

The predictive model understands the Captcha images. The better your training set is the better AI will be at deciphering the Captcha images. Companies using AI and human intelligence get better results. Now imagine using AI to tag your product images. You save tons of resources compared to manually doing these.

We understand your AI projects.

FarmOut Central is here to help. Our data processing team understands the needs of Big Data Analytics and AI. We do ETL (Extract, transform, load) processes for you. ETL activities include data cleansing, image tagging, outlier detection and fixing missing values.  This leaves more time for your analytics team to work on your predictive models.