Artificial Intelligence applied to the identification and automatic classification of MotoGP images

Within the projects carried out in the area of Artificial Intelligence, the Pasiona team has just developed a prototype for automatic identification and classification of teams in MotoGP.

A model trained with deep learning, using a large dataset of processed images to learn and achieve high accuracy in classifications and predictions. The system tracks the videos recorded in the MotoGP championships and is able to identify the teams protagonists in each of the cuts, in an automatic way and without the need for human intervention.

*Rights to video images: Dorna Sports

A model of this type is capable of generating considerable savings in time and money by automating the cuts of videos to be edited and published, thanks to the identification of images the sponsors of the teams can immediately receive all the images of the same at the end of the race, with the intention that they can assess the impact of their sponsorship. The project, born from a collaboration with Insight, opens a door to productivity and effectiveness for any company that provides multimedia material and especially to sponsors in sports events or any other large event.

“Working with Pasiona in this proof of concept has followed the same line as in previous opportunities: the treatment and closeness with its human team makes inter-business borders blur and we work as a single team,” says Javier Menéndez, Disruptive Solutions Manager at Insight. “This point was very important to work with Pasiona, but the fundamental element was the knowledge of talent and its pro-active way of acting” he adds, to highlight that “in these cases in which the choice of tools cannot be left in the hands of “trial and error”, the proactivity in proposing simple solutions to complex problems and the knowledge of the spectrum of technologies and models available to attack each casuistry made the PoC would flow quickly and efficiently,” he concludes.

The prototype posed different fronts and, from a convolutional neural network supported by the Tensorflow framework, artificial intelligence recognizes patterns and details in the images to increase accuracy and confidence in classifications. In addition, the model has the implementation of OpenCV to clean, segment, frame and display images. It also presents a tracking algorithm that helps you not to lose the identification of the teams once located. Thanks to the technologies involved such as Tensorflow, OpenCV and deep learning, after much work, the resulting prototype allows to clearly identify the different teams with an accuracy of 90%.

We bring you closer to the future (already present) with the infinite possibilities offered by Artificial Intelligence With a specialized and up-to-date team, we are prepared to find the solution you need. Shall we talk?

Contact us