Εκπαίδευση και Ανάλυση μοντέλων βαθιάς μάθησης για αναγνώριση κορονοϊού μέσα από αξονικές ακτινογραφίες με την χρήση Transfer learning
Date Issued
October 6, 2022
Type
Πτυχιακή Εργασία
Abstract
Coronavirus disease since December 2019 has significantly affected millions of people. Given the effect this disease has on the pulmonary systems of humans, chest radiographic imaging is now a necessity for monitoring the disease and preventing further deaths. Multiple studies showed that Deep Learning models can achieve great results in automatic diagnosis of COVID-19. Although, the number of available chest radiographs regarding COVID-19 is limited, hence these models have not been evaluated on data that’s new and unknown to the model. Regarding this thesis, 5 deep learning models were analyzed and evaluated with the aim of the identifying COVID-19 from chest X-Ray images. The dataset that was used is the largest one available publicly by the time of writing this thesis. Furthermore, the dataset is split into a train set for training each model, then there’s the validation set for validating the performance during training and then there’s the test set for evaluation on data that was not processed by each model. The models that were used are ResNet50, ResNet101, DenseNet121, DenseNet169 and InceptionV3 using Transfer Learning. All models were able to achieve accuracy of 93% and above. ResNet101 was the best performing model, whose accuracy reached 95.7%, the loss was at 0.13, Precision reached 95.7% and lastly Recall reached 95.6%.
Subjects