Skip to content

Latest commit

 

History

History
11 lines (7 loc) · 780 Bytes

README.md

File metadata and controls

11 lines (7 loc) · 780 Bytes

Lightweight model for the heavy dataset RAISE-LPBF

This repo is dedicated to my research on finding a lightweight model to reconstruct the laser parameters of the RAISE-LBPF dataset. Most of the architectures/models I worked on are omitted as well as the Neural Architecture Search (NAS) code.

More information on Makebench.

In brief, I tested many architectures to determine which one is the more appropriate for our case. One architecture was outperforming the other one, the 3D-CNN based models. I then managed to use Neural Architecture Search (NAS) to found the best 3D-CNN based model.

Below are the NAS results. NAS results