GLOBAL AI DATA TRAINING
From Collection to Delivery
To help a client build an app for the visually impaired, we sourced and delivered annotated image data of everyday items—captured in cluttered, real-world settings to reflect how people actually live and see.
From Blurred to Found: Empowering Vision Through AI
App development for medical conditions
Visually impaired people often have trouble finding items that are among other things. Whether that’s clutter on a desk or a kitchen counter with many different appliances and dishes, for example. To help people struggling with this, one of our clients developed an app that can scan and find what they’re looking for. Because this client would need very specific data collection and annotation to make the app a success, they came to us. We’ve built a reputation for delivering on these kinds of projects.
With our experience in both data collection and tagging for AI training, we were able to set up the parameters of the project by working closely with the client and learning exactly what would be needed. This wasn’t just about taking a picture of a cup and labeling it cup so the AI would be able to identify every cup the app was shown.
There was more to it.
The Challenge: Where Did I Put That Pen?
The client told us the app would only be successful if someone could find what they were looking for, even if it was in among clutter. For example, if we tasked our people to photograph a pen, it would have to be a pen on a desk with papers, a cup, a stapler, and file folders; in other words, clutter. From there, we would have to tag the pen in among the clutter – a natural environment – and annotate it.
Same for a kitchen counter. Finding a spoon on the counter among other items. Or a plate in a sink with cups and cutlery.
We were tasked to collect images for doors, windows, stairs, mailboxes, and common household items that a person might have trouble identifying clearly because of their impaired vision.
Data Collection: Training Requires Variety
We started out with a list of all the items the client wanted us to include in the first round of images, and then we set out recruiting people to photograph these images, including in different light conditions, which would alter what an image might look like.
A door or window at sunrise will look different at dusk, especially to a person who has trouble seeing. So the app needed to be very clear about what it was identifying. Once we had those parameters set, we began collecting images. datamundi.ai/ai-data-services AI Data Services
Post-Processing | Is That a Pen or a Paperclip?
Tagging and annotating proved to be a unique challenge at times. If we had trouble identifying a pen in the clutter, we could well imagine how hard it would be for a visually impaired person. This drove us to make sure we got as many useful, tagged, and annotated images as we could to help make the app a success.
Once we had what we knew to be a clean, tagged, and annotated dataset, we entered it into the training model.
The Result | Find what you’re looking for
With our expertly curated dataset, our client was able to successfully build out their app to help visually impaired persons identify what they’re looking at, or help them find what they’re looking for on a desk, a table, a kitchen counter, etc. Knowing that we were able to provide exactly what the client was looking for is one of DATAmundi’s strengths. With the large network of experts we can tap, we take on these challenges knowing we can deliver.
Our Data Services
We specialize in collecting and preparing a wide range of data types essential for AI development, including text data for Natural Language Processing (NLP), speech and audio, image, video, and multimodal data that combines multiple formats.
Learn more about our services on our Data Services page.
Or contact us below to see how you can be our next success story.
