Deep learning methods for 3D object recognition on meshes

Date

2024-9-03

Author

Akgül, Burak

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

312
views

205
downloads

Cite This

Recognition of objects is an essential step in many computer vision applications. An actively studied problem in this domain is the recognition of three-dimensional (3D) mesh models. Recent studies, especially those using machine learning techniques, have achieved remarkable accuracies in recognizing meshes, as obtained through evaluation on established datasets. In this thesis, we propose a machine learning model to recognize the object represented by a given mesh. Based on two-dimensional depth and volumetric data of the mesh, our approach involves a deep convolutional neural network, a novel symmetric difference operation, and significant data augmentation. By involving pre-trained models and an ensemble method, we further increase our model's accuracy. Our results on the ModelNet10 dataset ranks fairly high, especially among voxel-based methods.

Subject Keywords

3D mesh, Object recognition, Voxel map, Depth map

URI

https://hdl.handle.net/11511/111292

Collections

Graduate School of Natural and Applied Sciences, Thesis