Intel issues end-of-life notice for RealSense Lidar
therobotreport.com
therobotreport.com
For us it was the SR305 we were interested in as it had the specs we needed at the right price point. We had a meeting with the RealSense team about a year and a half ago and they strongly hinted at the SR305 being EOL soon which made it very hard for us to commit to RealSense for the project we were working on.
I'm quite surprised about this news though. Back then they suggested that LIDAR and stereo cameras would be their focus going forward. I can only assume this must be a recent move given what we were told back then and that the L515 is relatively new.
Huge shame about their facial auth products too. My understanding is that these are used in quite a few products, but they also use structed light tech so I assume that's why that decision was made.
It's worth noting that the L515 isn't replaceable by the D455 in many use cases. The D455 is good, but doesn't have the same kind of range and precision the L515 has. It's depth data is also far nosier, as is often the case with stereo depth cameras.
Does anyone have any experience with the TI mmWave sensors[1] for AMR collision avoidance?
The resolution and reliability was not as good as comparable devices in the price range like Microsoft's Azure Kinect DK.
Their software didn't work as well as what Microsoft cooked up, and it always felt like they put out this product and then just forgot about it.
And you’re still stuck with USB C on Kinect, so it’s not like it lets you escape the consumer-grade reliability.
If there are other devices with the same performance as the realsense lidar at a lower price, we’d all love to hear about them!
The people who designed USB-C were probably lazy people sitting in office chairs all day and never got out enough to see what real consumers do.
I break about a USB-C cable a week. Never happened with headphone jacks, barrel connectors, IEC power connectors, and the like -- those are real consumer grade stuff, especially IEC.
The connector design itself is shit, and the rubberized housing of the connector isn't mechanically supported as any self-respecting consumer-grade cable should be, all torque transfers to the PCB and that's horrible design.
There's no reason the torque has to transfer to the PCB. That's just poor design. My phone's USB-C port is sleeved by the chassis for example, and so is the one in my GaN charger. There is zero torque transfer to the PCB, the chassis or frame takes all of the load.
You can't rely on the cable to cancel the torque anyways, rubber yields and the normal force at the housing will pull the connector out slightly allowing for torque to still be applied to the connector, you have no choice than to brave either the female, male, or both ports to the chassis in some way.
Then there’s stuff built for public spaces. ATMs, books in the public library, turnstyles at the metro, etc. All designed to be in close contact with people who either just don’t care or are actively trying to damage it.
TBH I don’t know what the grade below consumer-grade would be.
There’s a type of consumer electronics available on Wish/AliExpress/etc., which is designed for people who want to appear at a glance to be using a certain expensive device, for status-signalling reasons. The devices only have to be real enough to allow you to pretend to be using them without drawing suspicion, and only have to be rugged enough to last until the device being mimicked isn’t fashionable to have any more (at which point you get a new knock-off aping the new cool thing.)
Surprisingly, these devices do work! They wouldn't be convincing otherwise. (How can you seem to be using the best new phone if you have to pull out a different phone to check your text messages? How can you seem to be playing a Switch or a PS4 if it doesn't turn on and show pictures and sound on the screen/a TV?) It's just that they work as terribly as you can imagine, given that costs are optimized to meet the minimum needs of someone playing pretend and no more.
Consumers are the ones that will actually have cables bitten, chewed, stomped, sat on by unfit, heavy civilian asses, jammed into wheels of office chairs, rammed into by strangers' hips in coffee shops when your connector hangs off the edge, dunked in cereal, splattered by kitchen grease, smashed into rocks at Yosemite, and shoved into bags while plugged in after a slow TSA checkpoint and needing to sprint as fast as possible to a departure gate.
Very different set of ruggedness requirements. Most cables of the 80s and 90s were completely fine in these civilian environments.
If the pro device breaks, the pro typically consumes a competitor's product. If the mil device breaks, the mil typically won't order 85,000 of them.
It is in the company's interest to ensure that the consumer device breaks.
The designers should take responsibility for it, and I pay them for that.
Headphone jacks worked great until they made goddamn headphones plug into USB-C as well and then they started breaking. Technology to withstand consumer lifestyles did exist until USB showed up.
Anyway, I’m not trying to get down on you - I was just pointing out that nothing is intended to last forever. I once considered buying a military grade reinforced laptop that was way under specced just because I wanted to take it wherever I wanted and be able to bang it around.
The truth is that electronics are easily damaged. Take fiber optic cables for instance. If you bend them ever so slightly too far they will stop working, because the throughput medium is glass. It’s just the nature of the medium.
Arguably, we are the most advanced machines on this planet, and yet if we slip and hit our our head, our machinery could easily be put out of commission. I think it’s fair to expect machinery to fail outside of its intended use case. Have you ever bent a piece of copper wire back and forth and feel it get warm before it breaks? That’s just the nature of the material. It is what it is, you know?
I’m not saying that you’re wrong to want or expect sturdier constructs - because they are often available for a price (using less abundant materials). I’m just saying that if you want that it’s probably not going to be consumer grade, and it’s not going to be cheap.
The majority of the world gets out and does things. Physically.
Also, as an active person myself I’ve never had a problem with USB-C; I don’t use my laptop out hiking and fishing, and I’m kind of confused who in the world is riding horses or bicycles and also fiddling with their consumer electronics at the same time.
And while maybe USB-C is less durable than headphone jacks, I can also say that wired headphones suck when you’re being active. I can’t even count how many times a door knob ripped out my earbuds and sent me into a fit of rage. I switched to Bluetooth and never looked back.
0 - Of course the world’s poor are active, but they’re also not the target consumers for USB-C devices.
Please don't dismiss things just because you don't do them.
I go on long bike rides with navigation on and that drains the battery quickly, so the needs to be constantly plugged into a portable battery pack, while in my pocket. This extremely common.
Consumer-grade has hints in the name that it is intended to be consumed and purchased again.
Business doesnt like making products that last for long life ad it means there is little chance of them buying another.
They can keep on trucking with R&D
So, intel made a custom processor for this that's really good at the correlations. The original one was the D4, and later they used their movidius chip. Both have lots of multiplier-accumulate silicon, so it can do the computations in parallel. Their architectures are also set up for convolution (which re-uses a lot of data) rather than random-processing (like a CPU does), so they could feed these math engines without a lot of data transfer -- this makes it more power efficient. You could do something similar in an FPGA, but dedicated silicon is going to be faster, cheaper, and use less power.
Of course, it works well as a tech demo and in controlled environments, and it's easy to understand and then brute-force with custom hardware. The problem is that it sucked the life out of better techniques which might actually work in the real world.
Google's core is search & advertising. But they are also in email, video streaming, messaging & video chat, cloud computing, mobile phones, gaming (Stadia), self-driving cars, drone delivery, quantum computing, and probably many other things that I can't think of.
Edit: Yes I know it’s technically Alphabet, but in practical terms, it’s Google
Lidar has high computational needs, which directly influences accuracy, and both spacial and time resolution.
Post-processing steps also have high computational needs, such as infering structure, do coregistration, handle point density, etc.
Also, practical applications often demand and depend on meeting constraints such as low power requirements and size.
If a company such as Intel managed to leverage their know-how to provide Lidar hardware that was competitive in both performance and price then they could as well develop a machine that prints money.
It'd be a bit silly for Intel to be successful in all of those spaces and not ask "and what would it look like to sell this part without bundling our CPU".
It followed the same familiar pattern, screw up the company over a long protracted acquisition, try to gain market share through bundling, all whilst driving the core talent ouf of the business by completely failing to invest. This is a big reason where Intel's problems came from. They looked to diversify rather than deliver.
> BMW Group, Intel, and Mobileye are joining forces to make self-driving vehicles and future mobility concepts become a reality. The three leaders from the automotive, technology and computer vision and machine learning industries are collaborating to bring solutions for highly and fully automated driving into series production by 2021.
Needless to say - a few years of toiling away and it turned out that this was not going to happen the way Intel had hoped. In fact if you go and look at BMW's website about autonomous driving it's almost comical that their list of acheivements stop the year before that deal was signed.
I'm not saying MobilEye aren't doing decent stuff (and they've certainly done a fantastic job of staying independent of Intel) but they've done nothing like what was expected of them. It's also important to remember that the bundling and strategic acquisitions Intel make often masks how successful they really are in a business - it'll often turn out the design win is more about bundling a bunch of services together rather than actually beating out competition.
Sure, this isn't living up to earlier estimates, but everyone in this field has been way behind early estimates, often by years. I think Mobileye revenue was up >100% last quarter? Obviously pandemic recovery helps, but even without it, 50+% seems likely.
If they can keep having failures like that, investors are going to be really happy with the outcome.
>. I think Mobileye revenue was up >100% last quarter?
As I say, they're doing fine. It's nothing like what they promised to do, and the company has restructured in a way that clearly indicates they'll be spun out in the next 24 months (like Mcafee). But the premise of the "big bets" wasn't shuffling at the edges, it was meant to be the future of Intel, and that's clearly failed at this point.
https://www.crn.com/news/components-peripherals/intel-says-i...
Does FLIR actually have a stereo solution? Back when they were Point Grey they sold the Bumblebee, but all of the stereo processing was done on the host machine.
An to put it in perspective, Intel's IOT group, which is (lol) mostly Point of Sale and MRI machines, does $3bn of revenue a year.
Intel is a big ship, and outside of new platforms (like smartphones were), there are few meaningful ways to diversify.
Hell, even if they bought NXP (one of the actual leaders in IOT), that would still only constitute 10% of their revenue.
And presumably they don’t think selling far fewer at a much higher price would do so either.
Of course it's management, but there is a particular kind of management that spawns around highly technical, highly profitable industries where engineers originally took the company to prominence.
Basically, engineers in management will be displaced by sociopathic MBAs in management that will use machiavellian social isolation combined with financing tricks.
In a way this goes back to a classic high school quote: popular people are popular because they spend all their time being popular. Nerds aren't popular because they spend their time learning.
So the "cool businesspeople" swoop in, isolate and eject the engineering-focused management with quick fix finance tricks and the usual middle management dirty pool.
Meanwhile the engineering-focused managers, invariably doing no-good-deed-goes-unpunished things like good engineering and good-for-the-companey... well, those people's days are numbered.
Then it devolves into a two-caste system: the engineers thrown enough money to keep the moneymaker ship on course and flowing, and the management caste partying on the decks.
This is not an environment conducive to moving into new markets.