2020/09/14 by Vincenzo Lomonaco, Lorenzo Pellegrini, Lomonaco, Vincenzo +27 · 1 citation
Computer Science · Medicine · #Artificial Intelligence (cs.AI) #COVID-19 diagnosis using AI #Computer Vision and Pattern Recognition (cs.CV) #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Multimodal Machine Learning Applications
paper · pdf · doi:10.48550/arxiv.2009.09929
openalex publication_date 2020/09/14 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
In the last few years, we have witnessed a renewed and fast-growing interest\nin continual learning with deep neural networks with the shared objective of\nmaking current AI systems more adaptive, efficient and autonomous. However,\ndespite the significant and undoubted progress of the field in addressing the\nissue of catastrophic forgetting, benchmarking different continual learning\napproaches is a difficult task by itself. In fact, given the proliferation of\ndifferent settings, training and evaluation protocols, metrics and\nnomenclature, it is often tricky to properly characterize a continual learning\nalgorithm, relate it to other solutions and gauge its real-world applicability.\nThe first Continual Learning in Computer Vision challenge held at CVPR in 2020\nhas been one of the first opportunities to evaluate different continual\nlearning algorithms on a common hardware with a large set of shared evaluation\nmetrics and 3 different settings based on the realistic CORe50 video benchmark.\nIn this paper, we report the main results of the competition, which counted\nmore than 79 teams registered, 11 finalists and 2300 in prizes. We also\nsummarize the winning approaches, current challenges and future research\ndirections.\n