51. SoftCTC -- Semi-Supervised Learning for Text Recognition using Soft Pseudo-Labels
- Author
-
Kišš, Martin, Hradiš, Michal, Beneš, Karel, Buchal, Petr, and Kula, Michal
- Subjects
Computer Science - Machine Learning ,Computer Science - Computer Vision and Pattern Recognition ,68T07, 68T10 - Abstract
This paper explores semi-supervised training for sequence tasks, such as Optical Character Recognition or Automatic Speech Recognition. We propose a novel loss function $\unicode{x2013}$ SoftCTC $\unicode{x2013}$ which is an extension of CTC allowing to consider multiple transcription variants at the same time. This allows to omit the confidence based filtering step which is otherwise a crucial component of pseudo-labeling approaches to semi-supervised learning. We demonstrate the effectiveness of our method on a challenging handwriting recognition task and conclude that SoftCTC matches the performance of a finely-tuned filtering based pipeline. We also evaluated SoftCTC in terms of computational efficiency, concluding that it is significantly more efficient than a na\"ive CTC-based approach for training on multiple transcription variants, and we make our GPU implementation public., Comment: 21 pages, 8 figures, 6 tables, accepted to International Journal on Document Analysis and Recognition (IJDAR)
- Published
- 2022