[2018.05.10] Edit Probability for Scene Text Recognition

[unizard]

CVPR2018

Institute: Hikvision URL: https://arxiv.org/pdf/1805.03384.pdf Keyword: SceneTextRecognition, Attention based Encoder-Decoder Network Interest: 2

Summary

• 기존연구의방법론
  • The existing methods usually employ a frame-wise maximal likelihood loss to optimize the models.
• 이런방법론이 당면한 문제
  • When we train the model, the misalignment between the ground truth strings and the attention’s output sequences of probability distribution, which is caused by missing or superfluous characters, will confuse and mislead the training process, and consequently make the training costly and degrade the recognition accuracy.
• 해결을위한방법론
  • To handle this problem, we propose a novel method called edit probability (EP) for scene text recognition.
• 방법론구체적으로?
  • EP tries to effectively estimate the probability of generating a string from the output sequence of probability distribution conditioned on the input image, while considering the possible occurrences of missing/superfluous characters.

페이퍼 잘쎴군. #준식이인턴가는회사 #연구잘하네ㅋㅋ

[unizard]

Fig. 1 provides examples to illustrate the phenomenon of missing and superfluous characters in training an attention based text recognition model on the ground truth “DOVE#”. Here, ‘#’ represents the End-Of-Sequence (EOS) symbol, which is commonly used in attention-based methods [7, 25, 32]. In Fig. 1 (a) and (b), the model may recognize the inputs as “DVE#” and “DOOVE#” respectively, based on the output pd sequences. Comparing against the gt “DOVE#”, it is natural to say that the former misses an ‘O’ and the latter has a superfluous ‘O’.

[raminrahimi6970]

is there implementation of this loss?