Classifying traffic signs using resnet architecture to optimise vision in autonomous vehicles
| dc.contributor.advisor | Kovács, László | |
| dc.contributor.author | Williams, Chidiebube | |
| dc.contributor.department | DE--Informatikai Kar | hu_HU |
| dc.date.accessioned | 2020-05-11T09:52:52Z | |
| dc.date.available | 2020-05-11T09:52:52Z | |
| dc.date.created | 2020-05-09 | |
| dc.description.abstract | The purpose of this study was to classify traffic signs the purpose of facilitate the performance of autonomous vehicles with respect to road signs. A popular German dataset known as GTSRB was used. The study used one of the state-of-the-art Neural Network architecture called the ResNet34 which is an already existing architecture and has been trained to work on over 1000 image classes. The architecture already has its weights so there was no need to put in my own weight. The originality of this study lies on trying to predict as many as 10 classes with a very high accuracy of more than 99%. This is also due to the fact that one of the methods used was transfer learning. During the study, the stochastic gradient decent was used and this was because of its ability to converge faster.The study was tested against so many other images, and an accuracy of 97% was achieved. The major challenge observed during the study is that training the dataset to classify about 10 classes or more was giving an accuracy as low as 60% which was also a big problem in letting the model predict an image in the test set. So, this was why the study was made not to focus on so many classes. | hu_HU |
| dc.description.course | Computer Science | hu_HU |
| dc.description.degree | egységes, osztatlan | hu_HU |
| dc.format.extent | 55 | hu_HU |
| dc.identifier.uri | http://hdl.handle.net/2437/286701 | |
| dc.language.iso | en | hu_HU |
| dc.subject | deep learning | hu_HU |
| dc.subject | machine learning | hu_HU |
| dc.subject | computer science | hu_HU |
| dc.subject | msc | hu_HU |
| dc.subject.dspace | DEENK Témalista::Informatika | hu_HU |
| dc.title | Classifying traffic signs using resnet architecture to optimise vision in autonomous vehicles | hu_HU |