"Continual Referring Expression Comprehension via Dual Modular Memoriza" by Heng Tao Shen, Cheng Chen et al.
Referring Expression Comprehension (REC) aims to localize an image region of a given object described by a natural-language expression. While promising performance has been demonstrated, existing REC algorithms make a strong assumption that training data feeding into a model are given upfront, which degrades its practicality for real-world scenarios. In this paper, we propose Continual Referring Expression Comprehension (CREC), a new setting for REC, where a model is learning on a stream of inco...
Expression Comprehension Continual Referring Expression Comprehension Dual Modular Memorization Continual Learning Lifelong Learning Eferring Expression Comprehension
Source: uow.edu.au