Institutional Repository of Vilnius University
Mažų objektų aptikimas infraraudonųjų spindulių vaizduose naudojant mašininio mokymosi metodus /
Abstract
dc:descriptionThis Bachelor’s final thesis involved comprehensive investigations and experiments to optimize the YOLOv5 neural network model for object detection in image data. The study focused on the effects of changes in the number of network layers and hyperparameters on the efficiency and accuracy of the model. Initially, the work involved adding more convolutional layers to improve the model’s ability to recognize objects. However, this alteration unexpectedly reduced the model’s accuracy, likely due to the presence of numerous objects with unclear features in the training data. The next experiment involved reducing the number of convolutional layers. The result showed that the model achieved higher accuracy more quickly, but the precise location of the object was determined less accurately than before. Further in the work, an input normalization (Batch Normalization) layer was added to accelerate the model’s learning process and enhance its stability. Yet again, this adjustment was not effective as the number of true positive predictions by the model decreased. In conclusion, the work revealed that the optimization of the neural network is a complex and challenging process requiring careful attention, experimentation, and well-coordinated modifications. Further research and experiments will need to be continued to achieve higher detection performance and accuracy.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Lakštinis, Justas,
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:ELABAETD192846790&prefLang=en_US