Hacker News Reader

Hi! Thanks for sharing this amazing work. I’m curious about the scalability and performance of PiLiDAR when deployed on large-scale outdoor datasets. Have you benchmarked it on datasets like SemanticKITTI or nuScenes? If so, could you share any insights on runtime, memory usage, and how well it generalizes beyond the indoor scenes used in your paper?

View on HN

I think you (or me, please correct me if that's the case) misunderstood something here - this is a diy lidar scanner for data acquisition - these datasets are mostly created using rgba cameras and the point clouds are later created with some post processing step.

So it's not a model for processing data but rather a hardware hack for having a real lidar - as in real depth data.

You can throw anything you like on it.

Reply on news.ycombinator.com