1. Program development for evaluating handwritten papers using neural networks
- Subjects
python ,machine learning ,education system ,анализ даннÑÑ ,маÑинное обÑÑение ,нейÑоннÑе ÑеÑи ,data mining ,neural networks ,обÑабоÑка изобÑажений ,ÑиÑÑема обÑÐ°Ð·Ð¾Ð²Ð°Ð½Ð¸Ñ ,image processing - Abstract
Рданной ÑабоÑе иÑÐ»ÐµÐ´Ð¾Ð²Ð°Ð½Ñ Ð¼ÐµÑÐ¾Ð´Ñ Ð¾Ð±ÑабоÑки изобÑажений Ð´Ð»Ñ ÑеализаÑии пÑогÑÐ°Ð¼Ð¼Ñ Ð¿ÑовеÑки ÑÑкопиÑнÑÑ ÑабоÑ. ÐÑи меÑÐ¾Ð´Ñ Ð±Ñли полноÑÑÑÑ Ð¿ÑоанализиÑÐ¾Ð²Ð°Ð½Ñ Ð¸ ÑлÑÑÑÐµÐ½Ñ Ð¿Ð¾Ð´ каждÑе ÑÐ¸Ð¿Ñ ÑÑловий. ÐÑли напиÑÐ°Ð½Ñ Ð°Ð»ÑÑеÑнаÑивнÑе ваÑианÑÑ Ð¸Ð·ÑÑеннÑÑ Ð°Ð»Ð³Ð¾ÑиÑмов, и в конÑе ÑоÑÑавлен иÑоговÑй ваÑÐ¸Ð°Ð½Ñ Ð¿ÑедобÑабокÑи. Ðомимо меÑодов ÑабоÑÑ Ñ Ð¸Ð·Ð¾Ð±ÑажениÑми Ñакже бÑли ÑÐµÐ°Ð»Ð¸Ð·Ð¾Ð²Ð°Ð½Ñ Ð·Ð°Ð´Ð°Ñи маÑинного обÑÑениÑ, в ÑаÑÑноÑÑи бÑли изÑÑÐµÐ½Ñ Ð¼ÐµÑÐ¾Ð´Ñ ÑаÑÐ¿Ð¾Ð·Ð½Ð°Ð²Ð½Ð¸Ñ ÑекÑÑа. ÐÐ»Ñ ÑÑого ÑÑпеÑно ÑазÑабоÑано неÑколÑко моделей, к одной из Ð½Ð¸Ñ Ð±Ñл ÑоÑÑавлен даÑаÑеÑ. Ð ÑеализаÑии Ñакже бÑл ÑазÑабоÑан алгоÑиÑм пÑовеÑки и пÑÐ¾Ð³Ð°Ð¼Ð¼Ð½Ð°Ñ Ð¾Ð±Ð¾Ð»Ð¾Ñка в виде полÑзоваÑелÑÑкого инÑеÑÑейÑа. ÐоÑле бÑло пÑоведено ÑеÑÑиÑование и апÑобаÑиÑ. РиÑоге бÑли полÑÑÐµÐ½Ñ ÑезÑлÑÑаÑÑ, пÑоанализиÑовав коÑоÑÑе бÑл ÑоÑÑавлен векÑÐ¾Ñ Ð´Ð°Ð»ÑнейÑего ÑазвиÑиÑ., In this work, the methods of image processing for the implementation of the program of verification of handwritten works are investigated. These methods have been fully analyzed and improved for each type of condition. Alternative versions of the algorithms were written, and the final version of the preprocessing was compiled at the end. In addition to the methods of working with images, machine learning tasks were also implemented, in particular, text recognition methods were researched. For this purpose, several models have been successfully developed, and a dataset has been created for one of them. The implementation also developed a verification algorithm and a software shell in the form of a user interface. After that, testing and approbation was carried out. Finally, the results were obtained, after analyzing which the vector of further development was compiled.
- Published
- 2021
- Full Text
- View/download PDF