Attention in CV

1.      Attention in CV

1.1.      CHAM:actionrecognition using convolutional hierarchical attention model.⁵

Cnn+多层LSTM，soft attention（指可导，hard attention使用强化学习，但是强化学习方差较大）

1.2.      Recursiverecurrent nets with attention modeling for OCR in the wild.⁶

  Recursive是结构上的循环（节省空间增加深度，类似与ISTN），recurrent是时间上的循环（多层使关注不同任务）。

1.3.      Show,attend and tell: neural image caption generation with visual attention.⁷

  Image caption任务。

1.4.      Recurrentmodels of visual attention.⁸

  鼻祖。

5.     Yan,S., Smith, J. S., Lu, W. & Zhang, B. CHAM: action recognition usingconvolutional hierarchical attention model.ArXiv Prepr. ArXiv170503146(2017).

6.     Lee,C.-Y. & Osindero, S. Recursive recurrent nets with attention modeling forOCR in the wild. inProceedings of the IEEE Conference on Computer Visionand Pattern Recognition 2231–2239 (2016).

7.     Xu,K. et al. Show, Attend and Tell: Neural Image Caption Generation withVisual Attention. in14, 77–81 (2015).

8.     Mnih,V., Heess, N., Graves, A. & others. Recurrent models of visual attention.inAdvances in neural information processing systems 2204–2212 (2014).