Image analysis via deep learning
One obstacle that people face when creating photo books is the choice of photos. Usually a person has thousands of pictures on their phone to choose from. To remedy this, the machine learning group at Luleå University of Technology will evaluate and use the latest knowledge in computer vision and, with the help of new deep learning techniques, analyze photos to come up with suggestions for clear photo books. The goal is to, with the help of Deep Learning and computer vision algorithms, select and group images for photo books with different themes, and to find patterns for how people are displayed.
The work is done in collaboration with the Skellefteå-based company Once Upon Publishing, whose business idea is to let its customers create photo books directly in their phones. The vision is that new users, within a few seconds after starting their app, will be presented with finished books, such as "Holiday in Rome" or "Family Yearbook". The suggestions follow different themes that are considered most relevant to the user. The customer should also be able to view a wider selection of photos from the selected events and adjust the selection of photos accordingly. For returning customers, the proposals should be based on previously created photo books.
Market-leading app for photo books
Once Upon provides an app for photo books that is designed and ordered easily from your mobile. Launched on the App Store and Google Play 2017, the app lets anyone create a photo book in just 20 minutes. The company's markets today are Western Europe, Australia and New Zealand. Once Upon had until recently about 25 employees, a figure the company expects to double in the coming year.
Linus Nilsson, Chief Experience Officer at Once Upon, says:
- Like so many others, we see great potential in AI technologies, but we also want to safeguard integrity and creative joy. A photo book is very personal and our hypothesis is that technology should help, but not take over. And it is important to clarify that we do not use our customers' images in the research project. At a later stage we will probably ask for volunteers, but so far we are working on our own pictures.
Growth in the AI industry
The work takes place within the framework of the project Applied AI DIH North at Luleå University of Technology. The project aims to create a strong innovation system for growth in the AI industry.
Michael Nilsson, project manager at the Center for Distance Spanning Technology, CDT, says:
- The knowledge that is now being built regarding image analysis of photo books can be applied to challenges in other industries. For example, real-time image analysis can be very useful for avoiding unnecessary stops in a production chain or for analyzing images in healthcare.
The Applied AI DIH North project will run for three years and is funded by the EU's regional development fund (Tillväxtverket), Luleå University of Technology, Luleå Municipality, Skellefteå Municipality and the Norrbotten Region. Read more at: http://www.aidih.se