Zhenmei Shi, Haoyang Fang, Yu-Wing Tai +1
The paper introduces DAWN, a dual augmented memory network that stores both target and background appearances using an attention‑guided LSTM memory, enabling robust unsupervised vi…