On methods of the transfer learning in the classification of the biomedical images

Jul 5, 2021, 3:45 PM
15m
407 or Online - https://jinr.webex.com/jinr/j.php?MTID=m573f9b30a298aa1fc397fb1a64a0fb4b

407 or Online - https://jinr.webex.com/jinr/j.php?MTID=m573f9b30a298aa1fc397fb1a64a0fb4b

9. Big data Analytics and Machine learning Big data Analytics and Machine learning.

Speaker

Prof. EUGENE SHCHETININ (Financial University under the Government of Russian the Federation)

Description

In this paper, computer studies of the effectiveness of the use of transfer learning methods for solving the problem of recognizing human brain tumors based on its MRI images are carried out. The deep convolutional networks VGG-16, ResNet-50, Inception_v3, and MobileNet_v2 were used as the basic models. Based on them, various strategies for training and fine-tuning models for recognizing brain tumors on a data set are implemented... Analysis of their performance indicators showed that the strategy for fine-tuning the ResNet50 model on an extended data set brought higher accuracy values, F1-metrics compared to other basic models. The best classification quality is achieved with transfer training on the VG 16 model with an accuracy of 95%.

Primary author

Prof. EUGENE SHCHETININ (Financial University under the Government of Russian the Federation)

Presentation materials

There are no materials yet.