Self-Supervised Learning to Reconstruct Dynamic Scenarios at Scale - NVIDIA DRIVE Labs Ep. 33

Опубликовано 28 февраля 2024, 20:16

#Autonomousvehicle #simulation is effective only if it can accurately reproduce the real world. The need for fidelity increases—and becomes more challenging to achieve—as scenarios become more dynamic and complex. Watch the latest #DRIVELabs to learn about EmerNeRF, a method for reconstructing dynamic driving scenarios. EmerNeRF builds upon NeRF (Neural Radiance Field) – which uses many 2D images to reconstruct 3D scenes – and extends it with self-supervised learning to accurately reconstruct dynamic scenarios and obtain labels for dynamic scenarios at scale.

Timestamp:
00:00:00 - Scaling diverse data in AV perception
00:00:27 - Introducing EmerNeRF, a self-supervised learning method
00:00:49 - Reconstructing scenarios into static, dynamic, and flow fields
00:01:40 - Lifting 2D foundation model features into 4D
00:02:15 - Using vision-language models for scene segmentations
00:02:40 - Dynamic scenario reconstruction at scale
00:03:02 - To learn more, visit our GitHub project page and blog

Project page: emernerf.github.io
Paper: arxiv.org/abs/2311.02077
Tech blog: developer.nvidia.com/blog/reco...

Watch the full series here: nvda.ws/3LsSgnH
Learn more about DRIVE Labs: nvda.ws/36r5c6t

Follow us on social:
Twitter: nvda.ws/3LRdkSs
LinkedIn: nvda.ws/3wI4kue
#NVIDIADRIVE