Self driving from VISION ONLY Teslas self driving progress by Andrej Karpathy ( Talk Analysis)
, tesla, selfdriving, karpathy Tesla is pushing the stateoftheart in full selfdriving, and interestingly, they explicitly switch from having multiple different sensors to a visiononly system. We discuss the highlights of Andrej Karpathy s talk about Tesla s FSD system, how to label petabytes of data, how to sample edgecases, how to train a neural network that has to work in realtime, and why moving to having only cameras is superior to multisensor approaches. OUTLINE: 0:00 Intro Overview 1:55 Current AutoBreaking system 3:20 Full SelfDriving from vision only 4:55 AutoLabelling for collecting data 8:45 How to get diverse data from edgecases 12:15 Neural network architecture 16:05 Tesla s inhouse supercomputer 17:00 Owning the whole pipeline 18:20 Example results from vision only 23:10 Conclusion Comments Links: TabNine Code Completion (Referral): YouTube:
|
|