Smart digital content platform

Smart Digital Content Platform

How to apply AI to facilitate searches in digital image repositories?

In this Digital Transformation Recipe, we will see how to use AI applied to image searches, to upgrade image retrieval times and reduce manual cataloging (indexing) work.
Picture of Veronica Meza

Veronica Meza

If you work in a communications or marketing department, or in a creative agency, you probably have thousands of images and you know that very often, finding or locating a digital resource can take more time than it should.

If you have a document management software, you will have metadata that will help you locate more easily the digital content you are looking for. But for those fields to have values, someone must have been in charge of filling them in. Unfortunately, we don’t always take the time to do this important work. In this digital transformation cookbook, we’ll look at how to use AI applied to image searches so that we can find images quickly while reducing the manual work of indexing.

As with most of our recipes, we’ll take it step by step – here’s the recipe!

How to use AI applied to image searches in 6 steps

1. Create a space

We will create a space to store our images. If you are not familiar with creating spaces in Athento, you can take a look at this video.

In my case, I have a space called “Under review”.

2. Create a form with a description field

In the Athento installation that our communication and marketing department works with, we have a form called Digital Asset. You can name it whatever name you like 😊, the important thing is that we will create a field called “description” for this form. This field is important because it will be the one in which we will use the AI applied to image searches. Using Artificial Intelligence we will complete this field automatically, without any human intervention. Once the AI completes it, we will be able to use it to find our images more easily.

I leave here the links to documentation so that you can perform this step:

Do not forget to activate the form in the space you created in step one (in case you have more than one space).

3. Upload your images to Athento

I already have 225 images uploaded in the Under review area. You will have to upload your images first before we can apply Artificial Intelligence to them.

To do massive uploads to Athento, we have two options:

  • Online upload: for uploading less than 20,000 documents.
  • Offline upload: this is done via API and is useful for uploading more than 20k.

Of course, you can also use the API upload, if you have your images in some application.

For our example, we are going to use the online upload. In this case, you just upload a CSV file with a column containing the title of each image and then drag and drop the files. Or, if what you want to do is just a test, directly from the drag & drop upload of the Dashboard, so that you don’t even have to indicate the title of the images.

I leave here the documentation of how the online bulk upload works and also other mechanisms for uploading files to Athento.

4. Connect to HugginFace

Athento offers integration with the Hugging Face platform to use AI models. Hugging Face is a community for collaborating, sharing and using machine learning models. Currently, Hugging Face offers more than 500,000 models.

You can follow this article to configure the integration between Athento and Hugging Face.

5. Activate the operation in space to automatically obtain the description of an image.

In the space where you have the images, you will have to activate the operation “Image captioning from document which basically uses a Hugging Face model to analyze the image and generate a textual description of it. Once we have our description, in the last step of this tutorial, we will be able to see how using AI applied to image searches, this description will make our life easier when searching and finding the images we want to use in the future.

If you are not familiar with automation tasks or operations in Athento, we recommend you to consult the documentation of this feature of Athento that will allow you to automate many tasks like this one. How to set up operations in a space in Athento

The configuration of this operation is very simple. Basically, in the case of this tutorial, we are going to indicate that the description returned by Hugging Face, we are going to save it in the “Description” field. The configuration would be as shown in the following image.

Another important aspect to consider is in which phase we activate the operation or automation task. In my case, I have configured it to run in the “On load” phase, that is, every time a new image is loaded in this space. But this will depend on your specific needs. Operations can be executed in Athento in multiple phases and events.


6. Try it!

Great! We are now ready to see the IA in action. In my case, I am going to run the operation on demand, manually, but as we saw in the previous step, operations can be run automatically at different stages of processing or events on a document.

Here is an animated gif in which you can see how we run the operation on a photo of our computer and the result that we get back.

IA aplicada a las búsquedas de imágenes

Now that we have our image with its automatic description we can see the power and benefit of AI applied to image searches. The description will make it much easier to find our image. For example, now if we search for the text “laptop” we can find our image.

And so much for our tutorial on AI applied to image searches! We hope you liked it and that you will put into practice this recipe for digital transformation with Athento.