Filters








3 Hits in 2.0 sec

FACLSTM: ConvLSTM with Focused Attention for Scene Text Recognition [article]

Qingqing Wang, Wenjing Jia, Xiangjian He, Yue Lu, Michael Blumenstein, Ye Huang
2019 arXiv   pre-print
, FACLSTM, i.e., Focused Attention ConvLSTM, where the spatial correlation of pixels is fully leveraged when performing sequential prediction with LSTM.  ...  In this paper, we argue that scene text recognition is essentially a spatiotemporal prediction problem for its 2-D image inputs, and propose a convolution LSTM (ConvLSTM)-based scene text recognizer, namely  ...  Figure 1 . 1 Challenging samples of scene text recognition. Figure 2 . 2 Current solutions for scene text recognition.  ... 
arXiv:1904.09405v1 fatcat:mk4lvdwmvvedjhiauuhngaqrrq

FACLSTM: ConvLSTM with focused attention for scene text recognition

Qingqing Wang, Ye Huang, Wenjing Jia, Xiangjian He, Michael Blumenstein, Shujing Lyu, Yue Lu
2020 Science China Information Sciences  
FACLSTM: ConvLSTM with focused attention for scene text recognition. Sci China Inf Sci, 2020, 63(2): 120103, https://doi.  ...  , FACLSTM, i.e., focused attention ConvLSTM, where the spatial correlation of pixels is fully leveraged when performing sequential prediction with LSTM.  ...  [34] for scene text detection. The original dataset is composed of 800000 scene text images, each with multiple word instances.  ... 
doi:10.1007/s11432-019-2713-1 fatcat:rjd3spur4zckxdzlsdejc7xfze

Special focus on deep learning for computer vision

Xiang Bai, Yanwei Pang, Guofeng Zhang
2020 Science China Information Sciences  
This general method is not only limited to gait-based age estimation, but also can be used for face-based age estimation. (3) In "FACLSTM: ConvLSTM with focused attention for scene text recognition", Wang  ...  They propose the ConvL-STM model for reading scene text from 2D space, by which attention mechanism and character center masks are further adopted for enhancing the recognition performance. (4) In "CGNet  ...  Cui et al. study the problem of localizing small objects with a single shot detector. Wang et al. propose to effectively leverage 3D blendshape information for facial expression recognition.  ... 
doi:10.1007/s11432-020-2766-x fatcat:2bmrqwendfgingrbbzo6v2gwea