University of Twente Student Theses
Enhancing Semantic Segmentation for Indoor Environments: Integrating Depth Information into Neural Networks
Velikov, Kristiyan (2023) Enhancing Semantic Segmentation for Indoor Environments: Integrating Depth Information into Neural Networks.
|
PDF
347kB |
| Abstract: | Image segmentation is part of computer vision tasks, with significant implications for the accurate identification and classification of indoor settings. Enhancing the performance of neural network-based image segmentation models can be achieved through depth information integration. This study moves beyond a survey to experimentally evaluate the integration of depth data within these models and pinpoint the most effective methods. The research delves into the primary sources of depth data and their incorporation into neural network models for image segmentation. Furthermore, this study probes the impact of depth information on the performance of DeepLabV3 Plus architecture, specifically with the implementation of a Shape-aware Convolution (ShapeConv) layer on the ResNext101 backbone. The research was conducted using the NYU Depth Dataset V2, with a critical focus on addressing the intricacies, challenges, and limitations inherent to depth information integration. In doing so, the study offers insights into the optimization of image segmentation models, particularly in the context of indoor environment analysis. |
| 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/96088 |
| 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