Skip to content

Latest commit

 

History

History
1 lines (1 loc) · 556 Bytes

README.md

File metadata and controls

1 lines (1 loc) · 556 Bytes

In this work, we address the lack of imagery by introducing additional constraints on the optimisation of theNeural Radiance Field by using depth information providedby a LIDAR sensor.From experiments and various ablationstudies we have conducted, we have concluded that usingthe constraints of a LIDAR sensor alone is not sufficient toachieve the desired results. In addition to depth constraints,we extend and experiment with the notions of extrapolatedLidar depth, feature loss, semantic loss, GAN-based loss and inverse-depth smoothing loss.