Scene understanding for unmanned vehicle using deep learning
In this dissertation, I learned some mainstream deep learning algorithms for scene understanding in the world. By comparing the performance of these algorithms on the PLACE365 dataset, finally proposed a method to use the Transformer layer as a pooling layer, and applied it to some mainstream deep l...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis-Master by Coursework |
| Language: | English |
| Published: |
Nanyang Technological University
2022
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/155393 |
| Summary: | In this dissertation, I learned some mainstream deep learning algorithms for scene understanding in the world. By comparing the performance of these algorithms on the PLACE365 dataset, finally proposed a method to use the Transformer layer as a pooling layer, and applied it to some mainstream deep learning networks for testing. By applying this method to the more classic neural network models such as Resnet18, Alexnet and VGG16, experiments on different datasets have shown that this method has a greater improvement in the accuracy of the model. |
|---|