AMLD EPFL 2024 marked a significant convergence of minds in the field of artificial intelligence (AI) and machine learning (ML), attracting professionals from diverse sectors to Lausanne, Switzerland, from March 23 to 26, 2024. Among the esteemed participants was Andrea Diecidue, who showcased pioneering work conducted by Politecnino di Milano – POLIMI – within the PERIVALLON project, particularly focusing on waste detection and localization efforts.
The conference, spanning four dynamic days, served as a hub for over 1500 attendees representing 40 countries. It featured a rich tapestry of activities including more than 450 speakers, 43 tracks, 28 workshops, keynote sessions, poster presentations, an exhibition, and startup pitches. AMLD EPFL 2024 offered a platform for attendees to delve into the latest advancements in AI and ML while fostering collaborative discussions and knowledge exchange.
Diecidue’s presentation underscored the integral role of POLIMI in developing cutting-edge tools capable of analyzing remote sensing images of vast areas, with a specific emphasis on identifying landfills and conducting comprehensive evaluations through drone inspections. Within the broader scope of the PERIVALLON project, Diecidue shed light on POLIMI’s multifaceted contributions, which include the development of AerialWaste, a dataset for waste detection, the development of OdinWeb, a web tool for annotating data and evaluating model performance, and the creation of a framework for training and evaluating deep learning models. Additionally, POLIMI established a pipeline for analyzing vast territories using the trained model. POLIMI’s ongoing research focuses on exploring new architectures and techniques for waste localization.
Furthermore, Diecidue highlighted ongoing research endeavors of POLIMI, as part of the PERIVALLON project. Specifically, POLIMI is continuing their research on binary and multi-class, multi-label classification on satellite images for waste detection. Additionally, they have started a new research path on drone data. Their objective is to identify specific targets and gauge the threat level of landfills, exploiting both 2D images and 3D point cloud data. Finally, they are constantly working on the AerialWaste dataset, providing more and more update.