From June 18 to 22, members from our partner Bruno Kessler Foundation (FBK) attended the Conference on Computer Vision and Pattern Recognition (CVPR), which stands as the leading yearly event for computer vision, encompassing the primary conference and multiple associated workshops. The conference was held in the Vancouver Convention Centre in Canada, gathering professionals worldwide and every aspect of the computer vision, machine learning and artificial intelligence industry. More than 10,000 CVPR attendees participated in CVPR in 2022. For this CVPR 2023, the focus has been on social networking, recruiting, learning, and reconnecting.
On behalf of AI-PRISM, Luigi Riz, Cristiano Saltori, Elisa Ricci, and Fabio Poiesi were privileged to present the paper “Novel Class Discovery for 3D Point Cloud Semantic Segmentation” as a poster. The paper primarily focuses on Novel Class Discovery (NCD) for semantic segmentation in 3D point cloud data. This topic previously had yet to be explored/researched, particularly in comparison to its 2D counterparts. The paper emphasised the unique challenges and the potential solutions in transitioning from 2D to 3D data, introducing a new method centred on online clustering and an evaluation protocol specific to point cloud data. The results, especially on the public datasets SemanticKITTI and SemanticPOSS, indicated a significant improvement over the baseline.
In the conference, the reception from the computer vision community, encompassing both research and industry professionals, was overwhelmingly positive. They had around 30 visitors to their poster, many of whom expressed that the idea was refreshingly novel. Several interactions with attendees indicated genuine interest in the approach, and potential applications were widely discussed.
Fabio Poiesi, Research Scientist at Kessler Foundation, expressed: “The overall experience at CVPR 2023 was unparalleled. Being surrounded by top-quality research presentations and cutting-edge discoveries, we felt our work was in an esteemed company. This opportunity has given us invaluable feedback and opened doors for future collaborations and research discussions. In conclusion, the presentation was a successful endeavour, and we look forward to further explorations arising from this experience. We are now in the phase of extending this work for a journal publication.”
The topic of the presented paper pertains to AI-based Perception, specifically the semantic understanding of 3D scenes. This work is crucial for AI-PRISM because semantic segmentation enables us to pinpoint objects within a scene and identify each object’s category, thereby facilitating the creation of high-level descriptors. Additionally, our participation in CVPR 2023 provided AI-PRISM with a distinctive platform to gain international acknowledgement in the computer vision community.
Overall, this esteemed conference highlighted European research excellence. Staying updated with the latest advancements ensures that our project aligns with current cutting-edge innovations, and the interactions at the conference can pave the way for potential international collaborations, infusing our project with varied perspectives. Furthermore, this platform enabled us to measure our work against global benchmarks, confirming our project’s competitive edge. The CVPR venue is a pivotal platform for amplifying our European project’s international impact and significance, promoting expansion and cooperation in the expansive realm of computer vision.
Do you want to learn more about AI-PRISM and the team behind the project? Subscribe to our newsletter and follow us on LinkedIn and Twitter, so you don’t miss a thing!