Decoded-ViT: A Vision Transformer Framework for Handwritten Digit String Recognition
Bibliographic record
Abstract
In the era of digitization, handwritten document recognition has several applications, like historical information preservation, postal address recognition, etc.The conservation and analysis of priceless cultural treasures depend heavily on the handwritten digit string recognition from historical documents.The prominent challenges in recognition are writing style variations, noise, distortions, and limited data.This paper suggests a novel method for overcoming the difficulties in reading complex, fading, and old handwritten documents that contain digit strings.The goal is to create a reliable and effective system that automatically recognizes digit strings from ancient manuscripts, helping to digitize records.Hence, this paper proposes a robust vision transformer framework to identify handwritten digit strings without segmentation of digits from uncleaned images of smaller datasets.The proposed method is a four-step procedure comprised of preprocessing, feature extraction through tokenization, recognition using the attention mechanism of a vision transformer, and outcome decoding using a beam search decoder.The performance of the proposed method is compared with the hybrid approach consisting of a Convolutional Neural Network and Long Short-Term Memory (CNN-LSTM).The proposed method achieved word accuracy of 56% with a loss below 0.6 in less time.The results show that the proposed model is a fast learner and can be used in real-time scenarios where results are expected in less time.The proposed deep learning model performance explanation is also discussed in this paper with the help of the Local Interpretable Model-agnostic Explanations (LIME) technique.The results of this study impact the digitization of postal services.The generalization of the proposed method by providing Software-as-a-Service (SaaS) for real-time applications is explored as a future research direction.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.004 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.019 | 0.009 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".