REXASI-PRO’s Post

🎥  The first year of the project told through a video made at the General Assembly. With a special focus on the three use cases and many open work perspectives!   🟣 🔵 🟠 🔴 

⭕ Our two-day General Assembly, held at Deutsches Forschungszentrum für Künstliche Intelligenz (DFKI) headquarters, concluded yesterday in #Bremen.   Thanks to all that attended the meeting. It was the precious occasion of intensive discussions, in plenary and in parallel sessions. 🧑💻👩💻 A profound and fruitful time of discussion ahead of the final year of the project. 📑 To all of us, happy continued work! 💪 For all our followers, the suggestion is to keep following us! 😉

🎙️ Vanessa Orani (Aitek S.p.A.) will present REXASI-PRO project on Friday at the headquarters of Confindustria Genova in an event organized by the latter with Digital Innovation Hub Liguria and Gruppo Dixet-Confindustria Genova.   📌 Coffeetech is an informal event, which stimulates the creation of relationships. The audience is heterogeneous, attended by companies, technology experts, researchers and students. The aim is also to create relationships and possible synergies. It will be possible to follow the event at the following link ➡️➡️ https://lnkd.in/dpdst6Pg To participate, you must register here ➡️➡️https://lnkd.in/dAnmTqNy

🚀 New Article on Our Website! 🌱🤖 REXASI-PRO is committed to developing a more environmentally-friendly and efficient AI. 🌍✨ We have just published a new on our website where we explain our approach to reducing the environmental impact of artificial intelligence through data reduction techniques. By training models with a carefully selected, representative subset of data, we achieve outstanding performance while significantly reducing CO2 emissions and speeding up training time. 🔍 In this case, we applied our strategy to a model for detecting people and people in wheelchairs, using the YOLO machine learning model. Our results show that this method delivers almost identical performance with much fewer resources. 📰 Read more about our approach and results in our website 👉 https://lnkd.in/dzvzCmJZ 📄 You can access our full paper published in Open Research Europe here: https://lnkd.in/dVfrJe6U 📹 We also invite you to watch our explanatory video, where we compare both models in action. Don’t miss it! https://lnkd.in/dvMr4g4V

Entitled “Implementation of safe AI” is the tutorial presented by Maurizio Mongelli (CNR-IEIIT) in Lecco at IEEE RTSI 2024. The study is also the outcome of ongoing results in REXASI-PRO and PRAESIIDIUM. If you would like to learn more about the content presented ➡️➡️ https://shorturl.at/sUvax

The REXASI-PRO project uses Machine Learning to control wheelchair motors using a Deep Neural Network that interprets sensor data. ♿ Would you like to know more? 🧐 A new in-depth study is now available on our website! 👇

📆 We are approaching our second General Assembly, which will be held next October 7 and 8 in Bremen, at Deutsches Forschungszentrum für Künstliche Intelligenz (DFKI) headquarters. ⭕ Meeting sessions to update ongoing activities and integration tests between the involved technical partners are some of the items on the agenda. To follow the developments of this new meeting opportunity, stay in touch with us! 😎

🟣 🔵 🟠 🔴 The REXASI-PRO approach for an environmental sustainability #TrustWorthyAICluster

Two publications related to the REXASI-PRO project have been accepted! 🤓   👉 In the first one, we discuss how to assess trustability of a LLM (like ChatGPT) to answer domain-specific queries, as in the situation where a wheelchair answers using its manual as a source. 👇 Sandra Mitrovic, Matteo Mazzola, Roberto Larcher, and Jerome Guzzi, "Assessing the Trustworthiness of Large Language Models on Domain-specific Questions”, EPIA Conference on Artificial Intelligence, 2024 (to appear)   👉 The second publication discusses how a ground robot (could be the smart wheelchair, but in our experiment we use different robots) can help a peer robot that has poor localization in the context of multi-robot system. 👇 Nicky Zimmerman, Alessandro Giusti, and Jerome Guzzi, "Resource-Aware Collaborative Monte Carlo Localization with Distribution Compression", International Conference on Intelligent Robots and Systems (IROS), 2024 (to appear)   Feedback from the scientific community continues for our project activities! 😎