Tamazight/Tifinagh-OCR-39K
收藏资源简介:
该数据集是一个全面的合成图像集合,包含39,101个样本,专为训练和评估Tifinagh脚本的OCR和视觉语言模型而设计。数据集具有多种字体、背景颜色和文本样式,采用矩形格式。每个样本包含图像文件路径、真实转录文本、背景颜色、文本颜色和唯一图像标识符等信息。适用于Tifinagh OCR训练和评估、Amazigh语言的文档理解、视觉语言模型验证以及跨不同排版样式的鲁棒性测试。
This dataset is a comprehensive collection of 39,101 synthetic images designed for training and evaluating OCR and vision-language models for the Tifinagh script. It features a wide variety of fonts, background colors, and text styles in a rectangular format. Each sample contains the image file path, ground-truth transcription text, background color, text color, and unique image identifier. Suitable for Tifinagh OCR training and evaluation, document understanding for Amazigh languages, vision-language model validation, and robustness testing across different typographic styles.




