PLATTER: A Page-Level Handwritten Text Recognition System for Indic Scripts

TL;DR

PLATTER addresses the lack of fair, end-to-end evaluation for page-level handwritten OCR on Indic scripts by decoupling the task into word-level HTD followed by HTR, enabling independent analysis of each stage. It advances a language-agnostic HTD model, benchmarks six HTR models (three CRNN, three Transformer) across ten Indic languages, and introduces CHIPS, a page-level dataset with detection and recognition labels, along with open-source code and trained models. The framework supports end-to-end OCR, single-page visualization, and cross-model latency/accuracy comparisons, revealing that CRNNs often yield robust end-to-end performance while certain Transformers excel in isolated WRR but may lag in end-to-end scenarios. Overall, PLATTER provides a reproducible benchmark and a practical end-to-end OCR workflow for Indic scripts, paving the way for consistent evaluation and future enhancements such as script-aware word detection and more robust language-agnostic recognition.

Abstract

In recent years, the field of Handwritten Text Recognition (HTR) has seen the emergence of various new models, each claiming to perform competitively better than the other in specific scenarios. However, making a fair comparison of these models is challenging due to inconsistent choices and diversity in test sets. Furthermore, recent advancements in HTR often fail to account for the diverse languages, especially Indic languages, likely due to the scarcity of relevant labeled datasets. Moreover, much of the previous work has focused primarily on character-level or word-level recognition, overlooking the crucial stage of Handwritten Text Detection (HTD) necessary for building a page-level end-to-end handwritten OCR pipeline. Through our paper, we address these gaps by making three pivotal contributions. Firstly, we present an end-to-end framework for Page-Level hAndwriTTen TExt Recognition (PLATTER) by treating it as a two-stage problem involving word-level HTD followed by HTR. This approach enables us to identify, assess, and address challenges in each stage independently. Secondly, we demonstrate the usage of PLATTER to measure the performance of our language-agnostic HTD model and present a consistent comparison of six trained HTR models on ten diverse Indic languages thereby encouraging consistent comparisons. Finally, we also release a Corpus of Handwritten Indic Scripts (CHIPS), a meticulously curated, page-level Indic handwritten OCR dataset labeled for both detection and recognition purposes. Additionally, we release our code and trained models, to encourage further contributions in this direction.

Figures (16)

• Figure 1: Limitations of current HTD and HTR methods used for OCR of Indic languages are shown in the top row. We depict the schematic pipeline of PLATTER for end-to-end page-level OCR below.
• Figure 3: Image pre-processing for generating word-level images to be incorporated in CHIPS
• Figure 4: Distribution of different word heights (in pixels) for words used in CHIPS pages
• Figure 6: Bounding boxes engulfing the words of a single Hindi page-level image produced by (L to R) ground truths, Tesseract, pre-trained DBNet, and our language agnostic fine-tuned DBNet respectively
• Figure 7: Validation loss vs epochs for the training of various HTR models for the Bengali language