Member Info for TheLongestShot

pm, here is the link to the press release:
https://www.iqep.com/media/press-releases/2021/iqe-reaches-key-milestones-with-iqdn-vcsel-technology-for-long-wavelength-vcsels-on-150-mm-gaas-substrates/

Before I include the text, the gist is that their technology allows lasers (light sources) with longer wavelength to be made more efficient. So it is not relevant to us as 850nm and 940nm are the specific wavelengths of near infra red that we use. Moving to even longer wavelengths is possible, but there are the efficiency problems on the emitter and receiver, you have to worry about transparency in glasses and plastics that may be used in sunglasses, the background level at those wavelengths in the car from the sun and other sources.

===
09 March 2021
IQE reaches key milestones with IQDN-VCSEL™ technology for long wavelength VCSELs on 150 mm GaAs substrates

IQE has achieved key power and reliability milestones for its IQDN-VCSEL™ technology for advanced sensing applications at longer wavelengths on 150 mm GaAs substrates
IQE's customers benefit as they can use the same equipment and processes that were developed for shorter wavelengths to achieve longer wavelengths
IQE is currently engaged in customer-funded development programmes with key Tier 1 customers to optimise this technology for qualification in future 3D sensing applications
IQE plc (AIM: IQE, "IQE" or the "Group"), the leading manufacturer of advanced semiconductor wafer products for the global semiconductor industry, is pleased to announce it has achieved key power and reliability milestones for its IQDN-VCSEL™ technology for advanced sensing applications at longer wavelengths (1100-1600 nm) on 150 mm GaAs substrates. IQE is currently supplying wafers to Tier 1 partners and collaborating with them to refine this technology for next generation advanced sensing applications.

The patent SYSTEMS AND METHODS FOR EYE-TRACKING DATA COLLECTION AND SHARING is a can of worms. It uses eye gaze data and historical patterns to guess what you will do, then it tells surrounding vehicles about its guesses, they then think about it and warn of unseen hazards, which get fed back to the driver.

So the first step is vague enough, because it depends on the context of where the vehicle is in its surroundings. If we only considered give way sign junctions or roundabouts, or when selecting reverse, the problem becomes more manageable, but less useful as there are other constraints on the problem space. So would you do all this processing just so that you might tell another vehicle about your guess assuming there is a compatible vehicle around to comment? Even then how is it going to visualise our vehicle plans, then respond with its identified risks in any meaningful way.

So if this level of processing and sensing was available, will cars do all of this altruistic processing? The level of processing is huge, modeling other vehicles intent based on guesses. This is where ee might be to fix the last risks when driving once we have solved the first 99.9999% of the driving problems. Fixing the next part of a 5 or 6 nines problem is hugely expensive

The Fourier Transform to find the eyes patent won't keep me awake at night. It is scoped to feed the eye position back to the HUD system so it's extra unnecessary code if there is a DMS on board.

The method of solving the problem abandons years of research and even open source solutions to "invent" a new method that required less thinking by the researcher because it used an existing 'Lego brick' of code (the Fourier Transform) because it was easy to think about, not because it was effective.

https://www.iihs.org/news/detail/iihs-pioneers-new-method-for-testing-driver-attention

Good to see on those CAT links that they are also selling a dedicated product called Guardian 2 for highway vehicles

Cerence - Oct 2020 $1bn, now $5bn

And remember - Cerence Look ( listening) win = Seeing win

https://www.linkedin.com/posts/david-mcgee-814b132_cerence-introduces-cerence-look-revolutionizing-activity-6767464688298074113-dODi
https://www.linkedin.com/posts/david-mcgee-814b132_cerence-introduces-cerence-look-revolutionizing-activity-6767464688298074113-dODi

Nice find JC,

This part looks interesting :

0027] One system for automatically controlling a flight path of an aircraft in response to an alert (disclosed in U.S. Pat. No. 10,102,773) includes at least one camera 16 and a face recognition processor 18, as shown in FIG. 2. The face recognition processor 18 receives facial image data from the camera 16 and processes that image data to determine whether the pilot is authorized to fly the aircraft or not. If not authorized, the face recognition processor 18 issues an activation signal to an automatic control processor 20 directing engagement of an automatic control system. In accordance with one embodiment, the face recognition processor 18 is located onboard the aircraft and communicates with a pilot facial image database 14 containing files of image data representing images of the faces of authorized pilots, which database is stored in computer memory onboard the aircraft. The face recognition processor 18 is programmed to compare the image data acquired by the camera 16 to the image data stored in the pilot facial image database 14. The face recognition processor 18 is further programmed to generate either a confirmation signal if the acquired image data matches image data in one of the stored files of image data of authorized pilots or the activation signal if the acquired image data does not match image data in any one of the stored files of image data of authorized pilots.

This patent is one of the foundations of our current and future success. It is already granted in the US.

The change is just our Patent Attorney updating/correcting his address

That is a great patent to put on your CV.

Essentially Qualcomm have to do what we do naturally, while still supporting our natural vision so that head mounted AR looks convincing without us needing to fill a bucket afterwards.

JC has picked up on the eye tracking part, but the meat of this patent is manipulating what the camera sees as the head scans across a scene to the next thing that is important to the wearer. We have evolved to not need to do the complicated maths, instead we rotate our head at various speeds and our eyes to stop at just the thing that we want to see. Along the way, our brain may close our eyes, or just not process the images. Our brains fill in the gaps with low resolution images captured before the movement so our brain doesn't panic that we can't see.

Of course when you put on a AR/VR helmet, it is requiring the headset to present the just the right images at the same time, but with minimal delay or our brains start to feel the lag and will get motion sickness. So foviated rendering and de-blurring and stretching or squashing the image are required to focus the processing on the necessary areas.

What I have written above doesn't necessarily appear in the patent, but it will help you to decode it!

Great article Aaroon, thanks for sharing

1 car sold followed by 13 cars sold.
It is not hard to get that growth if you have almost nothing to start with. It is called SOP
relax

https://twitter.com/seeingmachines/status/1359019349978345472?s=09 #TuesdayTips: Be Extra Cautious at Night.Always exercise ‘extra’ caution at night, especially in tight manoeuvring situations. #driversafety #innovation #SeeingMachines #DriverMonitoringSystem https://t.co/F5s3WqEN1M Here the focus is on 'extra' which must mean they are using the 'outside' of the car as the focus. Could be that they are making sure that you are using your windows and mirrors, not just the reversing camera. Or checking that you are watching the areas of most risk of a contract with people or things.

From our Friend Alex Roy

https://groundtruthautonomy.com/how-well-did-knight-rider-predict-autonomous-vehicles/

JC and I looked at this at the weekend, look back on our posts.
The patent shows that there is no eye tracking just shoulders, nose ears eye and mouth position so they can see if you are facing the potential problem. Kudos thought for the exterior modelling to identify potential hazards that is where most of the training data went.

Thanks Terry, Milford does a great introduction and deep dive into eye tracking. It does focus on head mounted trackers, so the maths and pixel counts are a little different from what Seeing Machines are working with. But a good overview

JC, when looking at the Nio ET7 last week I spotted something on the dash. An AI avatar that you can talk to in Chinese, it has an animated face and which it can turn in different directions.
Here it is in the car:
https://youtu.be/fhxi0BMAL0I

Can you remember who did the patent for this? I remember you posting it and me not being that impressed.

I see you have found it JC, but that product is based on Netradyne's IP. They are a firm part US and part Bangalore founded in 2015. They proudly boast 500 million miles of data - for the outside and inside. This is captured using various telematic units.

I dont doubt that this gives them plenty of data for identifying risks outside of the vehicle.

Now anyone skilled in the art knows that gaze detection is hard and that Seeing Machines has solved it and have the granted patents to prove it.

This patent shows that Netradyne's only have very loose gaze identification based on shoulders and ears and nose and eyes. That is their position in a body model, not where the eyes are actually looking based on reflected glints. But for their requirements that is sufficient - is the person facing the direction of the potential incident?

Now, let's pause and reconsider. Is the system detecting drowsiness or sleep? No.

Is the system providing accurate gaze direction to any other substation in the car? No.

Has this system integrated with Adas - ALKS, Route planning, or Emergency Braking? No.

This is purely an aftermarket system with a very short moment of opportunity before it is overtaken by NCAP requirements

Crazy Tesla again. Let's hope they tie it in with the patent that JC found to use eye tracking to determine the the direction the driver is looking. https://www.theverge.com/platform/amp/2021/1/29/22256504/teslas-model-s-x-redesign-automatic-shifting-prnd-gears?__twitter_impression=true

Well done, Grasshopper

Sorry, Baxter, I didn't scroll down far enough