"Russian Cursive Handwriting (RCH) Dataset"

"The Russian Cursive Handwriting (RCH) Dataset comprises 12,500 high-quality images of handwritten Russian words rendered in authentic cursive style, divided into 10,000 training and 2,500 test images reflecting an 80:20 split. The training set covers diverse topical regions including Central Russia, Saint Petersburg, Black Sea Coast, Arctic Region, Siberia, Caspian Region, Volga Region, Ural Mountains, North Caucasus, Kaliningrad Oblast, Sakhalin Oblast, Republic of Tatarstan, Ivanovo Oblast, Amur Oblast, Crimea Republic, Tomsk Oblast, Ryazan Oblast, Voronezh Oblast, and Smolensk Oblast, while the test set focuses on Altai Region, Lake Baikal Region, Southern Russia, Kamchatka Peninsula, and Yenisei River Basin. The dataset spans broad semantic categories such as geography, history, economy, industry, infrastructure, education, innovation, culture, religion, environment, food, tourism, sports, and famous people, offering rich linguistic and visual diversity. Images are provided in preprocessed formats (binarized, noise-reduced, resized to 128\u00d732 pixels) to support consistent OCR model training. This dataset addresses the scarcity of large-scale Russian cursive handwriting resources and enables development and benchmarking of deep learning models for optical character recognition under both large-data and few-shot learning scenarios. "