Projects

Studio 42

Videos to relive a personalized sports challenge for every athlete

Funding: IFAB for PMI

Enabling technology: Artificial Intelligence, Big Data Analytics, Geographic Information Systems (GIS), High Performance Computing, Image recognition, IoT

Sustainable Development Goals

The Jubatus project focuses on developing an innovative solution to improve the experience of participants in endurance sporting events, such as marathons and triathlons, which are growing in popularity in Italy. The problem identified is the lack of a way for athletes to relive and share their experience after the event. Market research revealed that the most desired format is a vertical reel ready within 48 hours of the event. The solution proposed by Studio 42 is structured around a three-stage operational process:

A first filming stage carried out by videographers with cameras actively tracking the athlete, a fleet of drones with auto-tracking for aerial footage, and fixed cameras at key points along the course;
A second stage for recognizing the athlete and the discipline based on identifying and distinctive elements;
A final stage dedicated to semi-automatic video content creation.

Studio 42 uses cutting-edge solutions such as AI 2.0 for the entire portion of the process dedicated to content analysis and creation (e.g., improving image quality), with a pioneering approach aimed at exploring new application areas. The project aims to deliver an innovative, scalable tool, which will require high-performance computing resources.

Objectives

The goal is to improve the existing auto-editing solution by modeling the identification of objective factors and the generation of subjective, context-constrained factors with a good degree of “exploration.” Improving the solution requires training the algorithm based on the skills and experience gained.

Initial challenge

After taking part in a sporting event, participants want to be able to see and share their own experience. The event itself is considered the culmination of a journey in which the athlete invests time, energy, and resources in long, demanding training sessions. Every athlete wants to remember, relive, and share those moments. Photography is widely used, while video footage is relatively rare, since it is much harder to capture each individual athlete on video than in a photo.

Solution

The system developed consists of two components: athlete recognition based on identifying elements, and semi-automatic reel editing with a high degree of product industrialization. Accuracy in recognizing individual athletes stands at 99.4%, although editing precision has not yet reached the same standard.

Benefits

Reduced editing and production times for increasingly professional, emotionally engaging, high-quality videos for each individual.
Creation of a video format suited to sharing on social media platforms.
Exploration of a growing technology, with the potential to generalize it to other businesses.

IFAB’s role

In collaboration with Jubatus, IFAB handled the technical development of the project. The Foundation’s contribution was organized across several fronts:

The starting point was reliable recognition of each individual participant based on identifying elements — a non-trivial problem in crowded settings with variable filming conditions.
Building on this, IFAB worked on formalizing the criteria for image selection, distinguishing between objective factors — measurable and repeatable — and subjective factors tied to the context of the event, which enrich the narrative and improve the perceived quality of the edit.
These elements together fed into the optimization of the existing auto-editing solution, integrated with feeds from active-tracking cameras, auto-tracking drones, and fixed cameras, and enhanced with image quality improvement algorithms.
The algorithm was then progressively trained to refine both recognition accuracy and the overall quality of the resulting reel.

The collaboration with Jubatus gave Fondazione IFAB the opportunity to apply computer vision and automatic editing techniques to an original use case, building specific expertise in visual identification in dynamic settings and in the automatic generation of personalized multimedia content.