Great video once again! There was one question I've had since the Tesla AI Day, and that is, how do they obtain labelled data in the vector space so that they can train the model? They tell us that the features obtained from all 8 cameras is then transformed into a 3D representation in vector space, but how do they obtain the labels to do so?
Indeed that wasn't shared! I would assume they first created a model with a lot of data they produced and use it as a pre-trained version to transform the data now
@@Mrbeginning123 Yes, that could be the case too. In Andrej Karpathy's talk in CVPR 2021, he did say that they have completely eliminated the use of LIDAR. So using past LIDAR data could be what they used.
Wow😍 Its the simplest explanation of very sophisticated system I've ever seen👏👏 Still few crashes happened. Level 5 autonomy is really hard to achieve.
Hahahaha Man! I was not expecting another heart from you but you give me, thank you so much. I was thinking like you would be busy reading another AI/ML/DL paper for the next video but I think you are now doing rest. If so then it's okay to take your time and rest is good for your health. Thank you so much.
It is not only important, but necessary! :) All thanks to you and yes, in fact i am currently at the gym between two exercises haha! You seem to already think this way, but please be sure to rest, too, and before you start feeling to need it!
watch this video make me understand how Tesla autopilot works. even i didn't know this before. and this video teach me how to say a model clear. like it
Humans are incredibly good at extracting relative motion from the environment. For example, given a human with normal vision, some 3d information can be extracted from single video simply by introducing a slight delay between what the right and right eyes see. In scenes with a lot of horizontal motion, the human brain will automatically starts up 3d calculations and maps the general image into that 3d model i.e. depth perception. It's actually quite stunning at times. It has a way of pulling you in and being there. Really trippy with old black n white film. (full disclosure: using sbs crosseyed technique.) Working on a player, but my coding skills not the greatest so . . it's taking a bit. It's also somewhat surprising that other senses (such as audio, or even smell, touch [like the road]) aren't payed any attention to. Audio ques such as sirens and nearby engine revs provide a more complete understanding of the total environment the agent is moving through. As always, you got me thinking again...🤔😎😉
Thank you for the feedback! It is in the complete opposite of what I usually receive. Could you help me oriented where I should get better at explaining these concepts please? I would gladly try to improve myself overall, but maybe you could provide more details that I could focus on? Thanks again for the constructive feedback! I appreciate you took the time to let me know, it doesn't happen that often! :)
Great breakdown with relevant links explaining what they used (CNNs, Transformers and RNNs) Thank you!
All thanks to you Alex! 🙌😊
Great video once again! There was one question I've had since the Tesla AI Day, and that is, how do they obtain labelled data in the vector space so that they can train the model? They tell us that the features obtained from all 8 cameras is then transformed into a 3D representation in vector space, but how do they obtain the labels to do so?
Indeed that wasn't shared! I would assume they first created a model with a lot of data they produced and use it as a pre-trained version to transform the data now
Teslas have been seen equipped with lidar sensors so maybe they used that data as labels?
@@Mrbeginning123 Yes, that could be the case too. In Andrej Karpathy's talk in CVPR 2021, he did say that they have completely eliminated the use of LIDAR. So using past LIDAR data could be what they used.
Best explanation of Tesla neural network that I have seen! Thank you!
Thank you so much! Glad you think so 🙌😊
*Do you think for fog, rainy where camera can't see much they should also rely on the radar?*
They don't have any radar on the Tesla cars!
Wow😍 Its the simplest explanation of very sophisticated system I've ever seen👏👏
Still few crashes happened. Level 5 autonomy is really hard to achieve.
Indeed it is extremely complicated! Thank you so much Jemish! 😊
very nice, man!
Thank you!
@@WhatsAI You deserve it.
Hahahaha Man! I was not expecting another heart from you but you give me, thank you so much. I was thinking like you would be busy reading another AI/ML/DL paper for the next video but I think you are now doing rest. If so then it's okay to take your time and rest is good for your health.
Thank you so much.
It is not only important, but necessary! :)
All thanks to you and yes, in fact i am currently at the gym between two exercises haha!
You seem to already think this way, but please be sure to rest, too, and before you start feeling to need it!
@@WhatsAI Huge Respect
watch this video make me understand how Tesla autopilot works. even i didn't know this before. and this video teach me how to say a model clear. like it
Thank you! Glad you liked it and it could have this effect! 😊
Humans are incredibly good at extracting relative motion from the environment. For example, given a human with normal vision, some 3d information can be extracted from single video simply by introducing a slight delay between what the right and right eyes see. In scenes with a lot of horizontal motion, the human brain will automatically starts up 3d calculations and maps the general image into that 3d model i.e. depth perception. It's actually quite stunning at times. It has a way of pulling you in and being there. Really trippy with old black n white film. (full disclosure: using sbs crosseyed technique.) Working on a player, but my coding skills not the greatest so . . it's taking a bit.
It's also somewhat surprising that other senses (such as audio, or even smell, touch [like the road]) aren't payed any attention to. Audio ques such as sirens and nearby engine revs provide a more complete understanding of the total environment the agent is moving through.
As always, you got me thinking again...🤔😎😉
Incomprehensible, and poorly explained.
Take more time and try to explain yourself.
Thank you for the feedback! It is in the complete opposite of what I usually receive. Could you help me oriented where I should get better at explaining these concepts please? I would gladly try to improve myself overall, but maybe you could provide more details that I could focus on?
Thanks again for the constructive feedback! I appreciate you took the time to let me know, it doesn't happen that often! :)