Application of Deep Learning Methods to Processing of Noisy Medical Video Data

Cells count become a challenging problem when the cells move in a continuous stream, and their boundaries are difficult for visual detection. To resolve this problem we modified the training and decision making processes using curriculum learning and multi-view predictions techniques, respectively. There are two main approaches to low-quality images processing. This research is related to the second approach which deals with robustness of machine learning (Matiisen et al., 2019; Radosavovic et al., 2018; Li et al., 2021; Bachman et al., 2019; Wenzel et al., 2020). Particulary, the research deals with curriculum learning (Wang et al., 2021) which is learning from simple tasks to complex ones, and multi-view post-processing which allows to make the final prediction based on multiple preliminary predictions.