CEU Electronic Theses and Dissertations, 2019
| Author | Paziczki, Peter |
|---|---|
| Title | Finding the most beneficial annotation strategy |
| Summary | I am working at a company that is developing an autonomous driving software, highly relying on neural networks when it comes to perception. I am part of a team that creates the labelled data for these neural networks and am facing questions, such as what frames to label so neural network performance can be as high as possible, considering the financial limits. Our objective was to find the optimal frame distance for semantic segmentation where given the limited resources the highest network performance can be achieved. The experiment we were conducting was about training a certain neural network several times and comparing the performance of each trainings to better understand what the key variables were. |
| Supervisor | Prekopcsák, Zoltan |
| Department | Economics MSc |
| Full text | https://www.etd.ceu.edu/2019/paziczki_peter.pdf |
Visit the CEU Library.
© 2007-2021, Central European University