Combining Synthetic and Real Data for Low-Resource Historical OCR: A Manchu Case Study

• URL: http://arxiv.org/abs/2609.11495v1
• Date: Thu, 10 Sep 2026 13:02:11 GMT
• Title: Combining Synthetic and Real Data for Low-Resource Historical OCR: A Manchu Case Study
• Abstract summary: Manchu, now critically endangered, was one of the principal languages of the Qing empire (1636-1912)<n>Previous work showed that vision-language models trained only on synthetic Manchu word images can reach 87.4% word accuracy on real Qing manuscripts and prints.<n>This study examines how synthetic and real historical training data should be combined for low-resource OCR.
• Score: 7.085654644528252
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Manchu, now critically endangered, was one of the principal languages of the Qing empire (1636-1912), and its extensive archival record is increasingly digitized but remains difficult to search and analyze at scale. Previous work showed that vision-language models (VLMs) trained only on synthetic Manchu word images can reach 87.4% word accuracy on real Qing manuscripts and prints, leaving a substantial synthetic-to-real gap. This study examines how synthetic and real historical training data should be combined for low-resource OCR. Using 60,000 synthetic and 20,306 real historical word images, we evaluate three pretrained VLMs and a compact convolutional recurrent neural network (CRNN) under four regimes: synthetic-only, real-only, joint synthetic-real, and sequential synthetic-to-real training, following a common checkpoint-selection and archival evaluation protocol. Introducing real training images raises the leading configurations to between 95.09% and 96.28% word accuracy, while no synthetic-only configuration exceeds 87.92%. Synthetic supplementation substantially improves all three VLMs, whereas its marginal effect for the CRNN is sensitive to the training objective. Joint and sequential training yield broadly similar archival accuracy under the tested practical pipelines. A compact CRNN also reaches the leading performance range once real images are available, showing that model scale alone does not determine recognition accuracy. Finally, complementary errors among strong recognizers allow voting to raise accuracy to 98.27% without additional training, while an eighteenth-century Manchu dictionary provides a principled rule for adjudicating disagreements.

Related papers

• Compensating for Scarce Historical Images in Cross-Domain Cultural Heritage Retrieval Using Synthetic Aging [0.0]
  This study investigates whether synthetically aged contemporary images can replace or complement missing historical data in bidirectional instance-level retrieval.<n>Mixed real-synthetic training is compared with real-only baselines using proportionally scaled and fixed 300-batch-per-epoch schedules.<n>The results indicate that genuine and synthetic observations are complementary.
  arXiv  Detail & Related papers  (2026-09-08T14:01:50Z)
• Design and Validation of a Lightweight 1D CNN for Affective Touch Classification in Soft Plush Companions [59.49967357689445]
  Soft, sensorized companions offer a physically safe and emotionally intuitive interface for socially assistive technologies.<n>This study presents a complete open-source dilated real-based framework for the development and validation of deep learning models for affective touch recognition in soft interactive companions.
  arXiv  Detail & Related papers  (2026-04-16T10:20:05Z)
• From Memorization to Creativity: LLM as a Designer of Novel Neural-Architectures [48.83701310501069]
  Large language models (LLMs) excel in program synthesis, yet their ability to autonomously navigate neural architecture design-balancing reliability, performance, and structural novelty--remains underexplored.<n>We address this by placing a code-oriented LLM within a closed-loop synthesis framework, analyzing its evolution over 22 supervised fine-tuning cycles.
  arXiv  Detail & Related papers  (2026-01-06T13:20:28Z)
• Stylized Synthetic Augmentation further improves Corruption Robustness [4.206961078715932]
  This paper proposes a training data augmentation pipeline that combines synthetic image data with neural style transfer.<n>We show that although applying style transfer on synthetic images degrades their quality with respect to the common Frechet Inception Distance metric, these images are surprisingly beneficial for model training.
  arXiv  Detail & Related papers  (2025-12-17T18:28:04Z)
• BlendCLIP: Bridging Synthetic and Real Domains for Zero-Shot 3D Object Classification with Multimodal Pretraining [2.400704807305413]
  Zero-shot 3D object classification is crucial for real-world applications like autonomous driving.<n>It is often hindered by a significant domain gap between the synthetic data used for training and the sparse, noisy LiDAR scans encountered in the real-world.<n>We introduce BlendCLIP, a multimodal pretraining framework that bridges this synthetic-to-real gap by strategically combining the strengths of both domains.
  arXiv  Detail & Related papers  (2025-10-21T03:08:27Z)
• Finetuning Vision-Language Models as OCR Systems for Low-Resource Languages: A Case Study of Manchu [0.0]
  Manchu, a critically endangered language, lacks effective OCR systems that can handle real-world historical documents.<n>This study develops high-performing OCR systems by fine-tuning three open-source vision-language models.<n>LLaMA-3.2-11B achieved exceptional performance with 98.3% word accuracy and 0.0024 character error rate on synthetic data.
  arXiv  Detail & Related papers  (2025-07-09T11:38:20Z)
• Merging synthetic and real embryo data for advanced AI predictions [69.07284335967019]
  We train two generative models using two datasets-one we created and made publicly available, and one existing public dataset-to generate synthetic embryo images at various cell stages.<n>These were combined with real images to train classification models for embryo cell stage prediction.<n>Our results demonstrate that incorporating synthetic images alongside real data improved classification performance, with the model achieving 97% accuracy compared to 94.5% when trained solely on real data.
  arXiv  Detail & Related papers  (2024-12-02T08:24:49Z)
• Learning from Synthetic Data for Visual Grounding [55.21937116752679]
  We show that SynGround can improve the localization capabilities of off-the-shelf vision-and-language models.<n>Data generated with SynGround improves the pointing game accuracy of a pretrained ALBEF and BLIP models by 4.81% and 17.11% absolute percentage points, respectively.
  arXiv  Detail & Related papers  (2024-03-20T17:59:43Z)
• UniBoost: Unsupervised Unimodal Pre-training for Boosting Zero-shot Vision-Language Tasks [60.46473247205654]
  Using large-scale unsupervised unimodal models as pre-training can enhance the zero-shot performance of image-text pair models. Our experiments show that unimodal pre-training outperforms state-of-the-art CLIP-based models.
  arXiv  Detail & Related papers  (2023-06-07T18:26:22Z)
• Robust deep learning for eye fundus images: Bridging real and synthetic data for enhancing generalization [0.8599177028761124]
  This work compares ten different GAN architectures to generate synthetic eye-fundus images with and without AMD. StyleGAN2 reached the lowest Frechet Inception Distance (166.17), and clinicians could not accurately differentiate between real and synthetic images. The accuracy rates were 82.8% for the test set and 81.3% for the STARE dataset, demonstrating the model's generalizability.
  arXiv  Detail & Related papers  (2022-03-25T18:42:20Z)
• Towards a Competitive End-to-End Speech Recognition for CHiME-6 Dinner Party Transcription [73.66530509749305]
  In this paper, we argue that, even in difficult cases, some end-to-end approaches show performance close to the hybrid baseline. We experimentally compare and analyze CTC-Attention versus RNN-Transducer approaches along with RNN versus Transformer architectures. Our best end-to-end model based on RNN-Transducer, together with improved beam search, reaches quality by only 3.8% WER abs. worse than the LF-MMI TDNN-F CHiME-6 Challenge baseline.
  arXiv  Detail & Related papers  (2020-04-22T19:08:33Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.