Buiten, Koen (2022) Multi-View 3D Object Recognition: Selecting the Best Sequences of Views. Master's Thesis / Essay, Artificial Intelligence.
|
Text
mAI_2022_BuitenK.pdf Download (3MB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
toestemming.pdf Restricted to Registered users only Download (130kB) |
Abstract
Most state-of-the-art object recognition systems make use of a multi-view recognition approach. In most approaches, viewpoints are selected from predefined viewpoint setups, like orthographic, orbit or hemisphere setups. The ability to have a more informed way of selecting viewpoints, could potentially improve the performance of a multi-view based system. For real-file robotic systems it has the additional benefit of travel reduction, because only the viewpoints which are most informative are visited. Previous studies have shown that 2D shape measures such as entropy are a good measure of determining the goodness of a view in terms of object recognition. We propose a next-best-view selection model based on the predicted depth-entropy of views in the neighborhood of the current viewpoint. The input for the model is the point cloud of the object, taken from the current view. The output is a three by five depth-entropy map. The model for predicting the entropy map is based on the well known PointNet model. We define entropy as Shanonn's entropy taken from the depth image of the corresponding view. Local extrema in the entropy map determine the next-best-view. In this thesis we compare our model with various baseline viewpoint selection methods. We have shown that our approach both increases the classification performance and decreases distance traveled between viewpoints. Furthermore, we have an indication that our model generalizes well when tested on unseen object classes.
| Item Type: | Thesis (Master's Thesis / Essay) |
|---|---|
| Supervisor name: | Mohades Kasaei, S.H. and Schomaker, L.R.B. |
| Degree programme: | Artificial Intelligence |
| Thesis type: | Master's Thesis / Essay |
| Language: | English |
| Date Deposited: | 29 Aug 2022 10:24 |
| Last Modified: | 29 Aug 2022 10:24 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/28514 |
Actions (login required)
 |
View Item |