Hidding, Lars (2024) User-guided, semi-supervised thin-section segmentation. Bachelor's Thesis, Computing Science.
|
Text
bCS2024HiddingL.pdf Download (12MB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
toestemming.pdf Restricted to Registered users only Download (453kB) |
Abstract
Thin section analysis is important in determining the physical properties of a rock. Manual analysis is a time-consuming process, one that can produce inconsistent results. Deep learning methods promise largely automated analysis procedures, yet such models require a lot of training data. In order to obtain this training data faster than can be achieved using manual analysis, one requires an automated approach. In this thesis we propose to automate most of this process by combining traditional image segmentation methods, while enabling the user to make corrections. The result is a novel image segmentation method and a software solution that will assist researchers in quickly performing thin section analysis. In order to keep the segmentation consistent, we record the performed actions. Finally, we assessed our results with a user study, measured the accuracy, and differences within accuracy using the Jaccard distance compared to a manually segmented version of the image. From the obtained results we conclude that the current achievable accuracy is subpar, user corrections were rarely used because most participants spent their time with global segmentation. However the method is, in our eyes, still promising, yet needs to be improved before it can be used in actual lab procedures.
| Item Type: | Thesis (Bachelor's Thesis) |
|---|---|
| Supervisor name: | Kehl, C. and Miocic, J.M. and Mulder, S.J. |
| Degree programme: | Computing Science |
| Thesis type: | Bachelor's Thesis |
| Language: | English |
| Date Deposited: | 23 Jul 2024 10:54 |
| Last Modified: | 23 Jul 2024 10:54 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/33589 |
Actions (login required)
 |
View Item |