Lidar or no-lidar

 This article is an excellent insight into a very nuanced problem. Collecting data from visible light, electromagnetic spectrum vs. Lidar-based frequencies results in different data sets. I tend to agree that the visual spectrum might be better. However, Lidar data can be treated as a backup or secondary source for a voting machine learning architecture. An alternative to sophisticated vision systems would be intelligent infrastructure that helps AIs resolve complicated problems.

It appears certain that the manual data labeling process is still a foundation for building training data for autopilots. However, Auto-labeling is increasingly possible because the amount of manually created labels allows for partial automation. Also, we are still dealing with the casual inference problems when the generalization power of #ML algorithms is insufficient to appropriately react to situations that are poorly covered in the training data.

#machinelearning #supervisedlearning #datalabeling #labeling #manuallabeling #dataengineering #deeplearning #radar #lidar #autonomous #autonomoussystems #casualinference #artificialintelligence #robotics #data #neuralnetworks #datascience #selfdrivingcars #algorithms #ai #automation #systemarchitecture

https://bdtechtalks.com/2021/06/28/tesla-computer-vision-autonomous-driving/