# On Calibration of Scene-Text Recognition Models

arXiv: https://arxiv.org/abs/2012.12643 · HTML: https://mlsanity.com/paper/2012.12643/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 23 Dec 2020.

## Built on by (most cited first, 9 of 9)

- [Document Understanding Dataset and Evaluation (DUDE)](https://mlsanity.com/paper/2305.08455/) (2023)
- [Training Uncertainty-Aware Classifiers with Conformalized Deep Learning](https://mlsanity.com/paper/2205.05878/) (2022)
- [CLIPTER: Looking at the Bigger Picture in Scene Text Recognition](https://mlsanity.com/paper/2301.07464/) (2023)
- [TextAdaIN: Paying Attention to Shortcut Learning in Text Recognizers](https://mlsanity.com/paper/2105.03906/) (2021)
- [Out-of-Vocabulary Challenge Report](https://mlsanity.com/paper/2209.06717/) (2022)
- [Enhancing Trustworthiness of Graph Neural Networks with Rank-Based Conformal Training](https://mlsanity.com/paper/2501.02767/) (2025)
- [Context-Aware Selective Label Smoothing for Calibrating Sequence Recognition Model](https://mlsanity.com/paper/2303.06946/) (2023)
- [A Human-in-the-Loop Label Error Detection Framework Applied to Arabic-Script HTR Datasets](https://mlsanity.com/paper/2601.16713/) (2026)
- [Perception and Semantic Aware Regularization for Sequential Confidence Calibration](https://mlsanity.com/paper/2305.19498/) (2023)

## Abstract (arXiv)

In this work, we study the problem of word-level confidence calibration for scene-text recognition (STR). Although the topic of confidence calibration has been an active research area for the last several decades, the case of structured and sequence prediction calibration has been scarcely explored. We analyze several recent STR methods and show that they are consistently overconfident. We then focus on the calibration of STR models on the word rather than the character level. In particular, we demonstrate that for attention based decoders, calibration of individual character predictions increases word-level calibration error compared to an uncalibrated model. In addition, we apply existing calibration methodologies as well as new sequence-based extensions to numerous STR models, demonstrating reduced calibration error by up to a factor of nearly 7. Finally, we show consistently improved accuracy results by applying our proposed sequence calibration method as a preprocessing step to beam-search.