University of Twente Student Theses
Test-Time Adaptation for Skin Lesion Classification
Bostelaar, Thomas (2024) Test-Time Adaptation for Skin Lesion Classification.
|
PDF
3MB |
| Abstract: | Deep learning models have shown good potential in the skin lesion classification task. In the case of a domain shift, where the test data comes from a different distribution than the training data, deep learning models struggle to perform. Dermoscopy images are taken under a wide range of different circumstances, and as a result distribution shifts can exist between different data sources. In this paper standard deep learning models are combined with test-time adaptation techniques to adapt to such domain shifts in the setting of binary skin lesion classification. By using a dataset split based on visual properties, models are confronted with a domain shift. Several test-time adaptation techniques are used in an online setup and compared to the unadapted version of the model. The performance of the models are analyzed and attention maps are used to better understand some of the performances. Based on the results, it cannot be concluded that test-time adaptation offer a stable improvement over standard deep learning models. |
| Item Type: | Essay (Bachelor) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 54 computer science |
| Programme: | Computer Science BSc (56964) |
| Link to this item: | https://purl.utwente.nl/essays/98382 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page