| Title |
3D objektų rekonstrukcija naudojant giliojo mokymosi algoritmus |
| Translation of Title |
3D object reconstruction using deep learning algorithms. |
| Authors |
Pocius, Algirdas |
| Full Text |
|
| Pages |
57 |
| Keywords [eng] |
deep learning ; 3D objects ; Pix2Vox ; voxel-based reconstruction ; data perturbation |
| Abstract [eng] |
Reconstructing 3D objects from a single two-dimensional image is a highly relevant task with a wide range of applications in virtual reality, autonomous systems and robotics. While much research focuses on modifying network architectures, relatively less attention has been given to data preparation strategies and training set sizes. The work presented in this paper analyses how a well-prepared training set, a rich source of multi-view data (e.g., ShapeNet) and network adaptation solutions can improve the accuracy of reconstructed 3D models. Results are evaluated using standard metrics such as Intersection-over-Union (IoU). |
| Dissertation Institution |
Kauno technologijos universitetas. |
| Type |
Master thesis |
| Language |
Lithuanian |
| Publication date |
2025 |
