Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

SEED: Semantics Enhanced Encoder-Decoder Framework for Scene Text Recognition

About

Scene text recognition is a hot research topic in computer vision. Recently, many recognition methods based on the encoder-decoder framework have been proposed, and they can handle scene texts of perspective distortion and curve shape. Nevertheless, they still face lots of challenges like image blur, uneven illumination, and incomplete characters. We argue that most encoder-decoder methods are based on local visual features without explicit global semantic information. In this work, we propose a semantics enhanced encoder-decoder framework to robustly recognize low-quality scene texts. The semantic information is used both in the encoder module for supervision and in the decoder module for initializing. In particular, the state-of-the art ASTER method is integrated into the proposed framework as an exemplar. Extensive experiments demonstrate that the proposed framework is more robust for low-quality text images, and achieves state-of-the-art results on several benchmark datasets.

Zhi Qiao, Yu Zhou, Dongbao Yang, Yucan Zhou, Weiping Wang• 2020

Related benchmarks

TaskDatasetResultRank
Scene Text RecognitionSVT (test)
Word Accuracy89.6
289
Scene Text RecognitionIIIT5K (test)
Word Accuracy93.8
244
Scene Text RecognitionIC15 (test)
Word Accuracy80
210
Scene Text RecognitionIC13 (test)
Word Accuracy92.8
207
Scene Text RecognitionSVTP (test)
Word Accuracy81.4
153
Scene Text RecognitionIIIT5K
Accuracy93.8
149
Scene Text RecognitionIC13, IC15, IIIT, SVT, SVTP, CUTE80 Average of 6 benchmarks (test)
Average Accuracy88.35
105
Scene Text RecognitionSVT 647 (test)
Accuracy89.6
101
Scene Text RecognitionCUTE 288 samples (test)
Word Accuracy83.6
98
Scene Text RecognitionCUTE
Accuracy83.6
92
Showing 10 of 37 rows

Other info

Code

Follow for update