S-LOR: Supervised Stream Learning for Object Recognition.
IbPRIA(2023)
摘要
In a stream learning scenario, where new data arrives at a slow pace, it is crucial to leverage new knowledge at the same rate without losing prior knowledge, and without assuming data stationarity. This scenario presents a significant challenge for incremental learning, particularly for tasks such as object recognition in video streams. In this paper, a novel approach is proposed that uses a set of weak classifiers that evolves into ensembles to enhance the generalization power of the system, as new video subsequences of the same instances are presented. We evaluate the efficiency of our approach and compare with state-of-the-art methods using a benchmark dataset. The code is available at https://github.com/vilaB/object_recognition .
更多查看译文
关键词
stream learning,object recognition
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要