IMOS Lab - EPFL

Our students from the Machine Learning for Predictive Maintenance Applications course recently had the unique opportunity to visit the FMHL pumped-storage power plant in Veytaux as part of their final project. This visit wasn’t just theoretical — it was hands-on, and our students will be working on real-world datasets directly from the plant owner Alpiq!   This collaboration gives our students invaluable insights into how predictive maintenance can optimize the performance of critical infrastructures, like hydropower plants, using machine learning. They tackle real challenges, working with data that directly impacts the energy sector, showcasing the power of AI in predictive maintenance.   A huge thank you to Alpiq and HYDRO Exploitation SA for partnering with us and providing this incredible learning opportunity! We look forward to seeing the amazing projects our students develop. #MachineLearning #PredictiveMaintenance #Hydropower #RealWorldData #Alpiq #HydroExploitation #EducationInAction #FutureOfEnergy

Congrats to Mariam Hassan on a successful presentation of her poster at the ECCV workshop on Neural Field Beyond Conventional Cameras (https://lnkd.in/eqTAhWMu). Lots of interest and lot of very impressing work on thermal cameras in this workshop! Her work: https://lnkd.in/ehgaKV9s

🎉 Exciting moments At the EPFL ENAC Research Day 2024, two of our PhD students: Chenghao Xu and Sergei G. presented their posters. Chenghao shared his work on applying neural scene techniques for estimating building characteristics, and Sergei shared his recent research on neural operator-based symbolic model.

Excited to go to Milano 🇮🇹 this week-end and next week for #ECCV2024 where I will be (co-)presenting three pieces of work: 📰 At the main conference: 𝑆𝑖𝑚𝑝𝑙𝑖𝑓𝑦𝑖𝑛𝑔 𝑆𝑜𝑢𝑟𝑐𝑒-𝐹𝑟𝑒𝑒 𝐷𝑜𝑚𝑎𝑖𝑛 𝐴𝑑𝑎𝑝𝑡𝑎𝑡𝑖𝑜𝑛 𝑓𝑜𝑟 𝑂𝑏𝑗𝑒𝑐𝑡 𝐷𝑒𝑡𝑒𝑐𝑡𝑖𝑜𝑛: 𝐸𝑓𝑓𝑒𝑐𝑡𝑖𝑣𝑒 𝑆𝑒𝑙𝑓-𝑇𝑟𝑎𝑖𝑛𝑖𝑛𝑔 𝑆𝑡𝑟𝑎𝑡𝑒𝑔𝑖𝑒𝑠 𝑎𝑛𝑑 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒 𝐼𝑛𝑠𝑖𝑔ℎ𝑡𝑠, with Yan Hao, Olga Fink. 🔗 paper: https://lnkd.in/dnZ3BD2b | code: https://lnkd.in/dRE_Xfj8 Come see my poster (# 44) on Wed 2 Oct 4:30 p.m. — 6:30 p.m. ! ❓ At the Workshop on Uncertainty Quantification for Computer Vision (https://lnkd.in/dcts55bS): 𝑈𝑛𝑐𝑒𝑟𝑡𝑎𝑖𝑛𝑡𝑦-𝐺𝑢𝑖𝑑𝑒𝑑 𝐴𝑙𝑖𝑔𝑛𝑚𝑒𝑛𝑡 𝑓𝑜𝑟 𝑈𝑛𝑠𝑢𝑝𝑒𝑟𝑣𝑖𝑠𝑒𝑑 𝐷𝑜𝑚𝑎𝑖𝑛 𝐴𝑑𝑎𝑝𝑡𝑎𝑡𝑖𝑜𝑛 𝑖𝑛 𝑅𝑒𝑔𝑟𝑒𝑠𝑠𝑖𝑜𝑛, with Ismail Nejjar, Gaetan Frusque, Olga Fink. 🔗 paper: https://lnkd.in/dYC_nzkR 📷 At the Workshop on Neural Fields Beyond Conventional Cameras (https://lnkd.in/dNYUa34t): 𝑇ℎ𝑒𝑟𝑚𝑜𝑁𝑒𝑅𝐹: 𝑀𝑢𝑙𝑡𝑖𝑚𝑜𝑑𝑎𝑙 𝑁𝑒𝑢𝑟𝑎𝑙 𝑅𝑎𝑑𝑖𝑎𝑛𝑐𝑒 𝐹𝑖𝑒𝑙𝑑𝑠 𝑓𝑜𝑟 𝑇ℎ𝑒𝑟𝑚𝑎𝑙 𝑁𝑜𝑣𝑒𝑙 𝑉𝑖𝑒𝑤 𝑆𝑦𝑛𝑡ℎ𝑒𝑠𝑖𝑠, with Mariam Hassan, Olga Fink, Malcolm Mielle, also contributing a new dataset ThermoScenes ! 🔗 paper: https://lnkd.in/d-3hC7WM | code: https://lnkd.in/ehru6kYE | dataset: https://lnkd.in/e-ZdEKAi #ECCV #ComputerVision #Research IMOS Lab - EPFL

Happy to share that our MultiOOD has been accepted by #NeurIPS2024 as a 𝗦𝗽𝗼𝘁𝗹𝗶𝗴𝗵𝘁! MultiOOD: Scaling Out-of-Distribution Detection for Multiple Modalities In this work, we introduce the first-of-its-kind benchmark for Multimodal OOD Detection, named MultiOOD, characterized by diverse dataset sizes and varying modality combinations. We highlight the significance of integrating more modalities for OOD detection on MultiOOD and introduce an Agree-to-Disagree (A2D) training algorithm, inspired by the observation of the Modality Prediction Discrepancy phenomenon. We also introduce a new outlier synthesis algorithm NP-Mix that explores broader feature spaces and complements A2D to strengthen the multimodal OOD detection performance. Many thanks to my collaborator Yue Zhao and my supervisors Eleni Chatzi and Olga Fink for their great help and support! paper: https://lnkd.in/dbrn4QFJ code: https://lnkd.in/dv-uEBjZ I will also be in Milan next week to present our work "Towards Multimodal Open-Set Domain Generalization and Adaptation through Self-supervision" at #ECCV2024. Come to my poster section on 3 Oct at 4:30 p.m. and have a talk! paper: https://lnkd.in/dk5jC7da code: https://lnkd.in/drvUSYHW #NeurIPS #ECCV #machinelearning #ai #OOD #multimodal

I’ll be talking at EPFL this Friday about optimal inspection, monitoring and maintenance. I will revise fundamental principles of predictive modeling and optimization of inspection and maintenance, discuss successes and limitations, and then look at current and future trends. Pass by if you happen to be in Lausanne!  https://lnkd.in/dhUQk_nq Thanks to Olga Fink for the invitation!

🌟 We’re excited to announce that we’ve started posting presentations from #IMC2024 on our YouTube channel! 🎥✨ Don’t miss out on the latest insights and discussions. Subscribe to stay updated! 📈🙌 Speaker: Dr. Saeid Hedayatrasa, Flanders Make Abstract: The vibrational characteristics of structural components contain valuable insights into their inherent mechanical properties and overall health status. Consequently, there's a pressing demand for efficient physics-based inversion algorithms. These algorithms must effectively reconstruct responses at unmeasured locations and/or identify unknown mechanical properties using a sparse set of noisy sensory data. Physics-informed neural networks (PINNs) offer promising solutions by seamlessly integrating the governing physics equations into their framework. Nonetheless, challenges arise due to the complex, multimodal, and multiscale nature of vibrational responses, along with the spectral bias of PINNs. In this talk we present a novel PINNs approach for efficient multimodal mapping of vibrations in a composite laminate and identification of its orthotropic elastic properties, given limited number of noisy simulation data points. https://lnkd.in/gBtnXNTN

I'll be at #ECCV2024 this year, where Mariam Hassan will be presenting her work on Thermal+RGB imaging using NeRF (https://lnkd.in/ehgaKV9s) at the workshop on Neural Fields Beyond Conventional Cameras(https://lnkd.in/eqTAhWMu). This work is a collaboration with Florent Forest and Olga Fink of the IMOS Lab - EPFL. Find the code on Github (https://lnkd.in/ehru6kYE), the dataset on Zenodo (https://lnkd.in/e-ZdEKAi), and the pdf on arxiv (https://lnkd.in/ei3QNBHz). Come talk to us!

🌟 Exciting PhD Opportunity: Physics-Informed Graph Neural Networks for Composite Space Structures 🌟 at the IMOS Lab - EPFL. This PhD position is part of the Marie Skłodowska-Curie Actions (MSCA) program. More details can be found here (you can submit the application there directly as well): https://lnkd.in/exGTgvEj

And that's a wrap! 🎉 The final day of our conference was a huge success, filled with insightful presentations, exciting discussions, and inspiring ideas. 1. Dr. Kai Hencken (ABB) hosting a panel discussion with vinay sharma (EPFL), Dr. Rafael Carrillo (CSEM), and Dr. Saeid Hedayatrasa (Flanders Make). 2. Dr. Florent Forest (EPFL) hosting a panel discussion with Prof. Sepideh Pashami (Halmstad University) and Mathias Pawlowsky (Planted Foods AG). 3. Prof. Ajith Kumar Parlikad (Cambridge University) discussing "Asset Management in the Digital Age". 4. Benjamin Regener (Kernkraftwerk Leibstadt AG) and Dr. Rialda Spahić (Equinor). 5. Prof. Olga Fink (EPFL) hosting the final panel discussion with Urs Gehric (SBB), Benjamin Regener (Kernkraftwerk Leibstadt AG), Vincent Cheriere (Airbus), and Henrik Göthberg (Dairdux). 6. Team members of the IMOS Lab - EPFL, organizers of the IMC 2024 conference. A big thank you to all the speakers, attendees, and organizers for making it a truly memorable event. We can’t wait to see where these new collaborations and innovations take us next! 🚀 #IMC2024 #Research #Innovation #Collaboration