Autonomous @ADASEngineer

@andrewjlingley I really think TOF can cover most usecases especially in highway scenarios, though instant velocity might be really beneficial in urban scenarios (instant velocity)

Thinking a lot about FMCW vs TOF - Can they coexist or is FMCW Lidar the future? FMCW: Range ✅ Resolution(FOV) (☑️) Comercials (☑️) (Single Chip!) Instant Velocity ✅ TOF: Range (☑️) Resolution(FOV) ✅ Commercials (☑️) Instant Velocity ❌ What do you think?

@lkstimm I guess you need to have twitter premium. I have premium on my other account so I might move it to this one as I do not really utilize it on the other

@lkstimm Great Idea - Also did not find yet a good community covering the space!

@lkstimm As a Tipp: what I have seen is that if you want to grow your followers you should engage a lot with bigger accounts and add value and increase your visibility

@lkstimm There are not many cooperations really utilizing X very effectively and I also would say that the linked in audience requires very different approach compared to X. X to LinkedIn is like Reddit to Quora - tick-tock styled shorts may not make it here but Memes and Content

@lkstimm I am mainly a consumer so what do I know! (btw I think your shorts content is great!) - Until now I am mainly a consumer and plan to add a bit more content this your for the sake of extending my view on automotive and sensing Technologies!

@lkstimm A crypto scam I guess? 😂 Alternatively I think X is a great platform to build a big, engaging community and then I would take it from there.

Certainly very interesting developments in the (mono) vision space for depth estimation! The question I have is: What is the false positive rate with those approaches and how reliable is it in general? Still fascinating how fast the vision space is moving!

AK @_akhaliq

2 years ago

TikTok presents Depth Anything Unleashing the Power of Large-Scale Unlabeled Data paper page: huggingface.co/papers/2401.10… demo: huggingface.co/spaces/LiheYou… Depth Anything is trained on 1.5M labeled images and 62M+ unlabeled images jointly, providing the most capable Monocular Depth

32 386 2K 600K 1K

It’s so hard to predict when full self driving is a reality. 90% success rate doesn’t cut it. 99% success rate doesn’t cut it. 99.9 success rate doesn’t cut it. And the cost to improve further are exponentially higher.

@wholemars Yea Cameras are certainly cheaper compared to LiDAR though I hope we can finally get rid of the misconception that automotive LiDAR are in the 10s of thousands… realistically we are by now in the range of 500€ per long range sensor and further improvements can be expected.

#Tesla #Cruise #Waymo #Mercedes

Working with HD maps for autonomous driving is not the worst idea. Yes it is not the end game, though it allows to figure out object detection, decision making and general reliability on public roads, without adding unnecessary risk and complexity by trying to solve all at ones.

Some of them might survive (Luminar? Innoviz?) but it will be hard. You can already see those companies looking into industrial and infrastructure sector to sell their products - not because the want, it’s because they must. #Lidar #adas #spac #AUTONOMOUS

3) Something important to look into are the press releases of those SPACs talking about billion dollar deals - Those are just dreams based on projected car sales and far away from reality. (I might go into detail another time)

The consolidation is and will be tough to watch in the automotive LiDAR market. All those SPAC companies will soon face the hard reality of the long development cycles and hard competition of the automotive marked. What does that mean?