9 ms·
Oyster: Towards Unsupervised Object Detection from Lidar Point Clouds
- ChuckNorris89 3y ago"No need for that, it can all be done using cameras because humans don't have LIDAR but drive with their eyes." /s
- williamcotton 3y agoWhen I’m driving I can determine the difference between a cardboard box and a plywood box. One of those objects I am perfectly fine to drive over. That’s just one of an infinity of embodied decisions I am able to make while driving. How does LIDAR help with that? Isn’t this a software problem?
- ChuckNorris89 3y agoThe LIDAR won't think the moon is a yellow traffic light. https://www.youtube.com/watch?v=7UF-S2czdCk https://www.youtube.com/watch?v=7UF-S2czdCk
- williamcotton 3y agoOk… so that doesn’t answer my question. It seems like the LIDAR based systems would be just as apt to make different kinds of mistakes.
- ChuckNorris89 3y agoYou misunderstood. It's not choosing either lidar or vision, it's that you need sensor fusion of both to make more accurate calls, neither of them is enough alone.
- williamcotton 3y agoBut there’s a cost to the software system when you add more inputs both in training and prediction and more-so when the inputs are in conflict! If you think about what really needs to happen to be able to tell of a box is cardboard or plywood you need a fully embodied system, ideally similar “raising” an android in a human environment.
- ChuckNorris89 3y ago>the inputs are in conflict! Better for safety to have conflicting inputs that require an arbiter (even human), than a system that's confidently wrong on a single input and drives you at full speed into a 'Wile E. Coyote' wall painted to look like a tunnel. [1] Since the moon landing, for safety, all aeronautics have been built around triple voting arbitration system where the correct answer is the one voted by the majority. This triple voting system has been adopted in many ECUs running safety critical operation in modern cars and we should demand it for self driving tech as well if we wish to have autonomous driving that's actually autonomous and also doesn't confidently try to kill you at times. >But there’s a cost to the software system when you add more inputs both in training and prediction Yes, not getting people killed, who are traveling in fast moving heavy machineries, is a very expensive problem to solve(ask the aero industry), welcome to the real world of actual engineering where safety must be taken seriously, and not brushed aside by overhyped SV fantasies of playing fast and loose with regulations and human lives for the sake of "disruption". [1] https://www.youtube.com/watch?v=4iWvedIhWjM https://www.youtube.com/watch?v=4iWvedIhWjM
- williamcotton 3y agoAll aeronautics require a triple voting arbitration system where the correct answer is the one voted by the majority. But those are not generally autonomous systems and if they are they operate within much simpler environments. In the case of multiple sensors and a voting mechanism the pilot is still in control and making the decisions. These are two very different kinds of systems! SV fantasies of playing fast and loose with hopium. Ah, so that’s it. Your arguments are based mainly on emotion and not engineering. You still haven’t explained to me how these systems can deal with the original problem I presented. EDIT: You'll notice that the commenter appealing to the populism of saying snarky things about Silicon Valley is receiving more encouragement from this forum than I am. I am very much trying to keep things even keeled and ChuckNorris is doing very much the opposite. Would everyone prefer that we just shoot emotionally charged hot takes back at each other with a minimum of engineering focused discussion?
- juliushuijnk 3y agoThe point is, if you say Lidar is essential, then how do you deal with fog, etc. Happy to be proven wrong, but my understanding is that the tech simply can't overcome these problems. So you either say 'driver takes over' in these cases, or you create software that uses video to figure out where stuff is. Since you need cams to read signs, you'll have that hardware and some software anyway. So if you have this software that can use video to figure out where things are, why do you need Lidar? The car needs to work safely without the Lidar. So including that in the sensor fusion should not be required. You can point out cases where software fails, but that does not disprove the fundamental reason for not using Lidar. It perhaps means that the software isn't good enough yet, but that does not seem fundamental for cases where humans wouldn't also fail.
- perfmode 3y agonothing is essential. everything is contributory.
- williamcotton 3y agoAnd many times less is more!
- ChuckNorris89 3y agoThis argument works in students building self driving robots for competitions on a budget, not in 2 tonne cars transporting your family.
- williamcotton 3y agoI am already thinking about this in terms of safety.
- ndsipa_pomu 3y agoLidar is capable of viewing through fog, whereas ordinary cameras won't be able to. https://www.photonics.com/Articles/Lidar_System_Delivers_Vision_Through_Clouds_and/a66169 https://www.photonics.com/Articles/Lidar_System_Delivers_Vis...
- speed_spread 3y agoMaybe LiDAR would realize the cardboard box is full of kittens.
- xoa 3y agoWhen I'm driving at night I cannot reliably determine the difference between a moose and an empty stretch of road because unlike deer moose do not have reflective eyes and stand tall enough that headlights will merely show a few easy to miss dark lines (the legs). One of these (empty road) I am perfectly fine to drive through. The other will result in bad day all around (assuming either side survives). Ie, I find people like you just, like, mindbogglingly ridiculous. Sure, no single sensor is going to magically solve everything. But we have tens of thousands of deaths and millions of casualties a year with humans at the wheel, and a sizable number of them are due to the limits of human visual spectrum input. The goal here shouldn't be to match human limits but to greatly exceed them. FLIR, lidar, radar, ultrasonics, all around vision in general, all should be part of the final package. Crashes, injuries and deaths are expensive to put it mildly. The idea that cameras are or ever could be good enough is absurd, and that was if they were as good as human eyes, which they aren't. No amount of software or compute power is going to deal with lack of quality input data, and the visible spectrum alone just can't provide that consistently without costs we have long since determined are unacceptable (have ultra high output flood lights around all vehicles). More sensors beyond human senses can both cover scenarios we can't and help fill in for individual sensors not being as high quality as might be ideal yet.
- williamcotton 3y agoWhat I find mindbogglingly ridiculous is how emotionally charged all of the "the system must include LIDAR" proponents are. It seems to suggest otherwise is perceived as nothing but a coldly calculated cost cutting measure. None of the arguments I come up against are anything but a surface level "well duh, intuitively more sensors equals more better" along with a litany of "you must like murdering babies". We do not have autonomous systems that can deal with the kinds of problems that human drivers currently deal with, which as I illustrated, is an infinity of embodied problems based on the entirety of our lives spent interacting with our physical environment. My intuitions tell me that in order to have a machine that can make human-like decisions that it needs to have that exact same kind of embodied experience, that is to have interacted with the physical environment in the same manner as a human, which suggests to me that perhaps an android is the way forward. Then once the autonomous system is extracted from the android and placed in a car it can be further trained on additional sensors, like how a human driver can take the input of LIDAR.
- the__alchemist 3y agoHearing is unable to detect colors, so I won't use it.
- williamcotton 3y agoWhat does this have to do with autonomous driving systems? Does a car just start driving itself once you've added enough sensory inputs?
- the__alchemist 3y agoIt's re this point: > When I’m driving I can determine the difference between a cardboard box and a plywood box... How does LIDAR help with that? Do you see the analogy?
- williamcotton 3y agoNo, I can't, because there's a lot more going on than just having recognized an object. How can we tell that it is cardboard? Or if we see someone throw a ball into the street, how can we get an idea of how much mass the ball has? Or any number of such scenarios... the answer is, we have a lifetime of experience of interacting with our environment. We've seen cardboard boxes, felt them, crushed them up and tossed them around. We've thrown balloons and rolled bowling balls. We have a learned understanding of both ourselves and the objects in our environment and how they behave.
- the__alchemist 3y ago> No, I can't, because there's a lot more going on than just having recognized an object. How can we tell that it is cardboard How can my ears tell that an object is red? Eyes are fantastic things.
- williamcotton 3y agoKnowing the color, shape, or sound that something makes doesn't mean anything without a lifetime of experiences to frame those sensory inputs and especially with regards to driving! You're being obtuse!
- marcinzm 3y agoI do not rely on just my eyes when driving and if I did I'd have been in at least one accident. I also rely on hearing and feel (vibrations, difficulty steering, car movement, etc.). Studies support this by indicating those who have hearing difficulties are more likely to get into accidents. I also rely on an organic computer far more powerful than anything you can shove into a car with modern technology and a pair of cameras with better dynamic range than anything on the market. Plus the ability to dynamically reposition these sensors by a couple feet. Once you can shove a human brain's worth of compute and memory into a car, along with an amazingly good set of 360 coverage stereo cameras, great stereo microphones, and multi accelerometers we can start talking about "LIDAR not mattering because the car can do what a human does and a human doesn't have LIDAR." Of course then you've got 20+ sensors that may disagree with each other anyways so LIDAR is probably going to mean fewer sensors overall.
- williamcotton 3y agoWhat I'm pointing out is that this is not merely an issue of sensors and that focusing on sensory input is not the way to solve the problems of embodied decision making. The focus should be on creating a system that can make these kinds of embodied decisions. Read my other comment in this thread related to androids.
- marcinzm 3y agoNothing says the only way to have a car drive as safely as an optimal human requires a human analogue. You're making a massive strong claim without anything to back it up. For example, humans can drive cars in video games fairly well from a third person perspective which clearly is not the same as sitting in a car.
- williamcotton 3y agoAre you offering up any concrete path towards dealing with the kinds of scenarios that I am presenting? And what does a human controlling a virtual car have to do with a machine controlling an actual car? The human can easily control the virtual car in the same manner as a real car precisely because of their embodied knowledge of cars, roads, and other parts of our experienced reality!
- marcinzm 3y ago>One of those objects I am perfectly fine to drive over. I recommend you Google "kid in cardboard box run over by car"
- ape4 3y agoOr less dramatic, there might be something heavy in there that you don't want to hit.
- throwaway173738 3y agoOr nails. Lots of nails.
- iainmerrick 3y agoHow do you know the cardboard box is OK to drive over? There might be something inside it.
- heavenlyblue 3y agoI don't think you can easily determine whether the cardboard box is empty or full of concrete though. Unless you can, then kudos to you.
- Scubabear68 3y agoThis could be problematic if there’s an anvil in that cardboard box. Just sayin’.
- judge2020 3y agoObject detection is solved; I'm not sure if everyone advocating for LIDAR is theorycrafting or just hasn't watched any FSD Beta videos, but all of the mistakes the system makes in the course of the drive seem to fall into the realm of unperformant decision-making, like predicting what bikes or pedestrians are going to do, or being too timid to enter a roundabout with another car also entering the roundabout to your left, or stopping in the middle of the road on a divided highway during an unprotected left turn to yield for right traffic (instead of positioning itself parallel to the traffic in the median in the first place).
- donkeyd 3y agoOr phantom braking because it thinks a weird shadow is actually an object. Or not realizing the difference between a reflection on a guard rail and a road line marking (because of low sun) causing it to directly aim for a guard rail (something I personally experienced). Sure, vision object detection is 'solved'.
- aidenn0 3y agoThe more nuanced argument is that LIDAR can't be required for driving if you want to drive in inclement weather, so any L5 driving system needs to be able to drive without it. Of course that's like saying I don't need my glasses to drive since in heavy rain my vision is equally impaired.
- sebastos 3y agoPeople have dug into a weird trench insisting LIDAR is absolutely necessary because Tesla (but really by association Elon Musk) are trying to do it without. The reality is that with or without LIDAR, the entire problem space is very far from solved, and the distance to close has a lot more to do with prediction, decision-making, and abstract representation. By the time these things have come far enough that full self driving cars are a reality, it's very likely vision will have progressed as well. The whole self-driving race is a leap of faith that there is some mid-point between "current state-of-the-art" and "full human-grade general intelligence" that is good enough to safely drive a car. Nobody knows whether that's actually true. So far, all we know is that general low-level sensorial awareness isn't a sufficient condition for tolerable driving. We all suspected this was the case, and so now comes the long tail of discovering just how much more behavior you have to add before you've sufficiently emulated human driving. These systems are so complicated and the interactions so numerous that the horizon is very foggy. It's POSSIBLE that you don't have an acceptable driver until you're basically doing human-level reasoning. Nobody knows. The point is that if you need a significantly deeper level of inference, prediction, reasoning, world state estimation, etc. etc., then the LIDAR will end up irrelevant, because the processing chain will be so sophisticated that camera-only input is easily sufficient for the sensing aspect. And if that's the case, camera-only is by far preferable because it is cheaper and simpler. Saying that you need LIDAR is just as unfounded as saying you don't. And it has the downside that you're doubling down on a more expensive technology that is demonstrably unneeded by a sufficiently sophisticated intelligence (i.e. humans).
- nradov 3y agoMercedes-Benz Drive Pilot is the only true self-driving technology available to individual vehicle buyers today, and it uses LIDAR. This is only a limited form of self driving and not human equivalent, but it does work at least in limited circumstances. https://media.mbusa.com/releases/mercedes-benz-worlds-first-automotive-company-to-certify-sae-level-3-system-for-us-market https://media.mbusa.com/releases/mercedes-benz-worlds-first-... So far no one has demonstrated an equivalent system without LIDAR. As you state, it might be possible but no one knows how or even has a clear path to doing it.
- codeka 3y ago
- noduerme 3y agoDoes object detection, as of now, not rely on both broad translational variance and then temporal consistency of tracks?
- q87b 3y ago> in self-driving scenes
- ImHereToVote 3y agoCan we get a printout of Oyster smiling?
- Alifatisk 3y agoIt should be spelled as OySTeR.
- porphyra 3y agotbh I am so annoyed by people spelling lidar as LiDAR. Nobody spells radar (which stands for Radio Detection And Ranging) as RaDAR...