For more than a decade, Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU) and Procter & Gam-ble (P&G) have maintained a close and productive collaboration. Two years ago, this partnership ex-panded into the domain of artificial intelligence and visual computing, with a particular focus on baby care products; most notably, diapers.
What initially appeared to be a niche application has revealed itself to be a rich source of challenging and intellectually stimulating problems. Over the past two years, we have successfully completed five contract research projects, several of which are now being implemented in production or product en-hancement. Traditionally, P&G has approached us with specific challenges, to which we proposed tar-geted solutions. Recent discussion, however, have surfaced a set of computer vision problems that resist straightforward resolution.
These tasks invite multiple methodological perspectives: classical pattern recognition and image pro-cessing, modern learning-based approaches, and, increasingly, foundation model-based strategies. Pre-liminary experiments suggest that foundation models can yield rapid progress, though they often lack the adaptability and transparency afforded by classical techniques. Conversely, traditional methods de-mand greater effort but provide full control over the solution space. This observation led us to conclude that a systematic exploration of diverse approaches would be most fruitful. Out of this realization emerged the idea of a joint hackathon.
FAU brings considerable experience in organizing such events. For instance, in December 2023 we hosted the Reproduce CVPR Hackathon in Erlangen, attracting around 30 participants. More recently, we have established a monthly mini-hackathon format for our doctoral and master’s students, each session yielding tangible outcomes for outreach (e.g., the Long Night of Science) or teaching. In our experience, hackathons are an ideal tool to explore a solution space, by leveraging the diversity in individuals’ exper-tise and creative approach.
The hackathon is being jointly organized by FAU and P&G, with all major decisions taken collaboratively. Both institutions will be equally represented during the event, with three senior and three junior staff members each providing guidance and support to participants. Problem selection, task definition, and data curation are being prepared jointly, ensuring a balanced and rigorous framework. Importantly, the hackathon will operate within P&G’s Open Innovation Crowd Sourcing strategy (very much of the same spirit of our planned hackathon), and all data and tasks will be non-confidential.
The central theme is the extraction of knowledge from diaper imagery. While some features (such as size or artwork) are relatively straightforward, the properties of interest lie in the visual characteristics of the various components: top sheet, absorbent core, back sheet, fastenings, and elastics. The complexity of these structures poses significant challenges, particularly in ensuring that solutions generalize across new designs and product variants.
Tasks will be predominantly multimodal, combining image data with textual or measurement-based de-scriptions. Participants will be selected to have a good mix of expertise, covering both classical computer vision as well as cutting-edge foundation models. At the same time, the organizing team, too, reflects that diversity, having explored both methodological directions in prior joint projects.
Selection of participants will follow a competitive process. Applications will open in December, with can-didates initially screened based on CVs. Shortlisted applicants will be given a small problem set, followed by an in-person interview. Final selection will take place in February. While FAU students will form the core of the participant pool, applications will be open internationally, advertised through academic net-works and social media.
Please note: This is just an information regarding events taking place at SSC; public attendance is therefore not possible due to the character of the retreat.
