Institutional Repository of Vilnius University
Įvairių nuotraukų iškraipymų įtakos vaizdų klasifikavimo uždavinyje tyrimas /
Abstract
dc:descriptionThis study examines how various image distortions affect the performance of modern image classification models. Although classification models are typically trained on high quality images, in real world scenarios, images may be of poor quality — blurred, noisy, poorly lit, or low in contrast. The goal of the research is to determine how different distortions impact the models’ ability to accurately classify such images. A small subset of the ImageNet dataset was used in the study, along with three modern classification model architectures: ”ViTB/16”, ”EfficientNetB0”, and ”ConvNeXtTiny”. The experiments revealed that some models are more resilient to certain types of distortions than others. Overall, ”ConvNeXtTiny” was found to be the most robust against various types of distortions. To sum up, image quality degradation has a significant impact on classification accuracy, and that choosing the right architecture can help mitigate such effects.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Juškevičius, Gustas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD239030858&prefLang=en_US