中国科学院自动化研究所机构知识库

Online learning is crucial to robust visual object track- ing as it can provide high discrimination power in the pres- ence of background distractors. However, there are two contradictory factors affecting its successful deployment on the real visual tracking platform: the discrimination issue due to the challenges in vanilla gradient descent, which does not guarantee good convergence; the robustness is- sue due to over-fitting resulting from excessive update with limited memory size (the oldest samples are discarded).

Despite many dedicated techniques proposed to some- how treat those issues, in this paper we take a new way to strike a compromise between them based on the recursive least-squares estimation (LSE) algorithm. After connect- ing each fully-connected layer with LSE separately via nor- mal equations, we further propose an improved mini-batch stochastic gradient descent algorithm for fully-connected network learning with memory retention in a recursive fash- ion. This characteristic can spontaneously reduce the risk of over-fitting resulting from catastrophic forgetting in ex- cessive online learning. Meanwhile, it can effectively im- prove convergence though the cost function is computed over all the training samples that the algorithm has ever seen. We realize this recursive LSE-aided online learning technique in the state-of-the-art RT-MDNet tracker, and the consistent improvements on four challenging benchmarks prove its efficiency without additional offline training and too much tedious work on parameter adjusting.