emredeveloper/turkish-ocr

土耳其OCR合成数据集，包含1000张带有土耳其文字的逼真图像，适用于光学字符识别。数据集包含70%手写文本和30%打印文本，支持全部土耳其字符。图像背景包括有线条的纸张、网格纸和纹理纸。数据集还应用了运动模糊、噪声、透视失真和墨迹瑕疵等增强技术。

Turkish OCR Synthetic Dataset containing 1000 realistic images with Turkish text, suitable for Optical Character Recognition. The dataset includes 70% handwritten text and 30% printed text, supporting all Turkish characters. Image backgrounds consist of lined paper, grid paper, and textured paper. The dataset also employs augmentations such as motion blur, noise, perspective distortion, and ink artifacts.