BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
Townyj
Ermahgerd
KiKi
Regular
TWEET by Edge Impulse just 8 minutes ago
www.aalto.fi
The new smart sensor uses embedded information to detect motion in a single video frame
Conventional sensors only capture a single moment in a frame, but the new sensor can read information about the past and use that to predict the future. Image: Hongwei Tan / Aalto University
A new bio-inspired sensor can recognise moving objects in a single frame from a video and successfully predict where they will move to. This smart sensor, described in a Nature Communications paper, will be a valuable tool in a range of fields, including dynamic vision sensing, automatic inspection, industrial process control, robotic guidance, and autonomous driving technology.
Current motion detection systems need many components and complex algorithms doing frame-by-frame analyses, which makes them inefficient and energy-intensive. Inspired by the human visual system, researchers at Aalto University have developed a new neuromorphic vision technology that integrates sensing, memory, and processing in a single device that can detect motion and predict trajectories.
At the core of their technology is an array of photomemristors, electrical devices that produce electric current in response to light. The current doesn’t immediately stop when the light is switched off. Instead, it decays gradually, which means that photomemristors can effectively ‘remember’ whether they’ve been exposed to light recently. As a result, a sensor made from an array of photomemristors doesn’t just record instantaneous information about a scene, like a camera does, but also includes a dynamic memory of the preceding instants.
‘The unique property of our technology is its ability to integrate a series of optical images in one frame,’ explains Hongwei Tan, the research fellow who led the study. ‘The information of each image is embedded in the following images as hidden information. In other words, the final frame in a video also has information about all the previous frames. That lets us detect motion earlier in the video by analysing only the final frame with a simple artificial neural network. The result is a compact and efficient sensing unit.’
To demonstrate the technology, the researchers used videos showing the letters of a word one at a time. Because all the words ended with the letter ‘E’, the final frame of all the videos looked similar. Conventional vision sensors couldn’t tell whether the ‘E’ on the screen had appeared after the other letters in ‘APPLE’ or ‘GRAPE’. But the photomemristor array could use hidden information in the final frame to infer which letters had preceded it and predict what the word was with nearly 100% accuracy.
In another test, the team showed the sensor videos of a simulated person moving at three different speeds. Not only was the system able to recognize motion by analysing a single frame, but it also correctly predicted the next frames.
The sensor is made of an array of photo memristors. Image: Hongwei Tan / Aalto University
Accurately detecting motion and predicting where an object will be are vital for self-driving technology and intelligent transport. Autonomous vehicles need accurate predictions of how cars, bikes, pedestrians, and other objects will move in order to guide their decisions. By adding a machine learning system to the photomemristor array, the researchers showed that their integrated system can predict future motion based on in-sensor processing of an all-informative frame.
‘Motion recognition and prediction by our compact in-sensor memory and computing solution provides new opportunities in autonomous robotics and human-machine interactions,’ says Professor Sebastiaan van Dijken. ‘The in-frame information that we attain in our system using photomemristors avoids redundant data flows, enabling energy-efficient decision-making in real time.’
A neuromorphic visual sensor can recognise moving objects and predict their path | Aalto University
The new smart sensor uses embedded information to detect motion in a single video frame
A neuromorphic visual sensor can recognise moving objects and predict their path
Published: 17.4.2023The new smart sensor uses embedded information to detect motion in a single video frame
Conventional sensors only capture a single moment in a frame, but the new sensor can read information about the past and use that to predict the future. Image: Hongwei Tan / Aalto University
A new bio-inspired sensor can recognise moving objects in a single frame from a video and successfully predict where they will move to. This smart sensor, described in a Nature Communications paper, will be a valuable tool in a range of fields, including dynamic vision sensing, automatic inspection, industrial process control, robotic guidance, and autonomous driving technology.
Current motion detection systems need many components and complex algorithms doing frame-by-frame analyses, which makes them inefficient and energy-intensive. Inspired by the human visual system, researchers at Aalto University have developed a new neuromorphic vision technology that integrates sensing, memory, and processing in a single device that can detect motion and predict trajectories.
At the core of their technology is an array of photomemristors, electrical devices that produce electric current in response to light. The current doesn’t immediately stop when the light is switched off. Instead, it decays gradually, which means that photomemristors can effectively ‘remember’ whether they’ve been exposed to light recently. As a result, a sensor made from an array of photomemristors doesn’t just record instantaneous information about a scene, like a camera does, but also includes a dynamic memory of the preceding instants.
‘The unique property of our technology is its ability to integrate a series of optical images in one frame,’ explains Hongwei Tan, the research fellow who led the study. ‘The information of each image is embedded in the following images as hidden information. In other words, the final frame in a video also has information about all the previous frames. That lets us detect motion earlier in the video by analysing only the final frame with a simple artificial neural network. The result is a compact and efficient sensing unit.’
To demonstrate the technology, the researchers used videos showing the letters of a word one at a time. Because all the words ended with the letter ‘E’, the final frame of all the videos looked similar. Conventional vision sensors couldn’t tell whether the ‘E’ on the screen had appeared after the other letters in ‘APPLE’ or ‘GRAPE’. But the photomemristor array could use hidden information in the final frame to infer which letters had preceded it and predict what the word was with nearly 100% accuracy.
In another test, the team showed the sensor videos of a simulated person moving at three different speeds. Not only was the system able to recognize motion by analysing a single frame, but it also correctly predicted the next frames.
The sensor is made of an array of photo memristors. Image: Hongwei Tan / Aalto University
Accurately detecting motion and predicting where an object will be are vital for self-driving technology and intelligent transport. Autonomous vehicles need accurate predictions of how cars, bikes, pedestrians, and other objects will move in order to guide their decisions. By adding a machine learning system to the photomemristor array, the researchers showed that their integrated system can predict future motion based on in-sensor processing of an all-informative frame.
‘Motion recognition and prediction by our compact in-sensor memory and computing solution provides new opportunities in autonomous robotics and human-machine interactions,’ says Professor Sebastiaan van Dijken. ‘The in-frame information that we attain in our system using photomemristors avoids redundant data flows, enabling energy-efficient decision-making in real time.’
FrederikSchack
Regular
Agree!
FrederikSchack
Regular
I also see the same picture, institutions buying, while the price has been declining.
Then we have to think supply/demand who's supplying the shares? Most likely the private investors.
It's a shame that private investors doesn't have stronger hands....
Frangipani
Top 20
Hi @Mercfan,
since you have Munich‘s Allianz Arena as your profile pic, I assume you are not only a Mercedes, but also an FC Bayern fan? (Who knows, you might be a BVB fan instead, posting this beautiful sunset shot as an ominous Götterdämmerung prediction for Dortmund‘s biggest Bundesliga rival…
) I just happened to read somewhere that Bayern München are planning a tour through Asia in July. Let’s assume they will be playing each of their three planned matches in Bangkok, Singapore and Tokyo at 8 pm local time. For fans in Germany wanting to watch a live TV broadcast, however, the kick-off will be at different times each match day (3 pm German time for the match in Bangkok, 2 pm for the one in Singapore and 1 pm for the match in Tokyo), as there is no one single Asian time zone. Neither is there one time zone for all of Australia.So when you are saying that Germany is 8 hours behind Australia, this statement of yours is only partly true. It is indeed, if you are currently in let’s say Port Douglas (Queensland) or Port Macquarie (New South Wales). However, if you are in Port Hedland (Western Australia) instead, Germany is currently only 6 hours behind, as WA is in the same time zone as Singapore. And to make things even more complicated, the time difference between the Ports of Hamburg or Rotterdam and Port Augusta in South Australia is 7.5 hours!
But the point I was trying to get across in my previous post is that those time zones stated in the press release @Opti had asked about were actually North American time zones, not Australian ones.
One hint was that the press release was published by Brainchip’s US office in Laguna Hills, CA. It is located in the Pacific Time zone, which is encompassing parts of western Canada, parts of the western United States (except Alaska and Hawaii) as well as the state of Baja California in Mexico. The largest city in the Eastern Time Zone, on the continent’s opposite Atlantic Coast, is New York City.
So another hint - as already mentioned in my previous post - was that only in the Americas (but not in Australia) would it make sense to differentiate between a Pacific (= West Coast) and an Eastern time zone with a three hour time difference between them. In Australia, on the other hand, the East Coast is the Pacific Coast!
So the live online event started at 9.30 am New York time (opening time of the NYSE/NASDAQ) / 6.30 am Los Angeles time, which in turn was 3.30 pm Central European Daylight Saving Time. And way past dinner time (for most people at least) anywhere in Australia & New Zealand. Hope this helps.
D
Deleted member 118
Guest
Maybe he has one of theseHi @Mercfan,
since you have Munich‘s Allianz Arena as your profile pic, I assume you are not only a Mercedes, but also an FC Bayern fan? (Who knows, you might be a BVB fan instead, posting this beautiful sunset shot as an ominous Götterdämmerung prediction for Dortmund‘s biggest Bundesliga rival…
So when you are saying that Germany is 8 hours behind Australia, this statement of yours is only partly true. It is indeed, if you are currently in let’s say Port Douglas (Queensland) or Port Macquarie (New South Wales). However, if you are in Port Hedland (Western Australia) instead, Germany is currently only 6 hours behind, as WA is in the same time zone as Singapore. And to make things even more complicated, the time difference between the Ports of Hamburg or Rotterdam and Port Augusta in South Australia is 7.5 hours!
But the point I was trying to get across in my previous post is that those time zones stated in the press release @Opti had asked about were actually North American time zones, not Australian ones.
View attachment 34475
One hint was that the press release was published by Brainchip’s US office in Laguna Hills, CA. It is located in the Pacific Time zone, which is encompassing parts of western Canada, parts of the western United States (except Alaska and Hawaii) as well as the state of Baja California in Mexico. The largest city in the Eastern Time Zone, on the continent’s opposite Atlantic Coast, is New York City.
So another hint - as already mentioned in my previous post - was that only in the Americas (but not in Australia) would it make sense to differentiate between a Pacific (= West Coast) and an Eastern time zone with a three hour time difference between them. In Australia, on the other hand, the East Coast is the Pacific Coast!
So the live online event started at 9.30 am New York time (opening time of the NYSE/NASDAQ) / 6.30 am Los Angeles time, which in turn was 3.30 pm Central European Daylight Saving Time. And way past dinner time (for most people at least) anywhere in Australia & New Zealand. Hope this helps.
FrederikSchack
Regular
I found Google very helpful to translate the time.
Just search:
"9:30 AM eastern to german time"
Raspberry Pi Receives Strategic Investment from Sony Semiconductor Solutions Corporation
<p>Establishing strategic partnership for the further enhancement of the edge AI solution development TOKYO and CAMBRIDGE, United Kingdom, April 12, 2023 /PRNewswire/ -- Sony Semiconductor Solutions...</p>
FrederikSchack
Regular
I had a Brainchip Facebook group that was almost disintegrating after Brainchip had been shooting up from nothing to almost 1 AUD momentarily in September 2020 and then retreated back most of the gains. Some were unhappy and started shooting at me.
I left the Brainchip group and started a new group called Fundamental Stock Analysis, the purpose was to collaboratively contribute to analysis of stocks. This group learnt me that almost everybody wants to look, but not contribute anything.
My understanding is that people in general don't care about the qualitative fundamentals, they just want money fast without doing anything significant. They don't understand the value of the company, because they don't understand the fundamentals. So, they only have the price of the company to rely on and when it falls, that's all they see and they panic.
I think that people to some degree consider the market perfect, so they assume that the share price is falling because others know better, so in their head a fall is equal to some horrible new truths being realized somewhere. I've been winning from a lot of opportunities because of the imperfections in the market, I know the market is far from perfect and that's where the opportunities are.
I also realized that other peoples laziness is my opportunity, I can find those stocks that have incredible potential and buy them before everybody else sees it and are piling into them. It takes a lot of reading, but it has definitely been worth it. Same when stocks get overhyped and overcrowded too early, then I leave.
I consider Brainchip an outstanding opportunity at the current price. The value of the company has increased tremendously since three years ago, Brainchip is an entirely different company than back then and the market is much more primed for AI now than back then, but we're close to the same price tag as back then.
---------------------------
Just to reiterate my reasoning behind my view on the value of the company from a previous post:
1) AI got super focus after ChatGPT, Dale-E, Stable Diffusion e.t.c. and the amount of companies that are making something similar is exploding and many more billions will be poured into AI. We're at the accelerating part of the technological S-curve in AI. That's super positive for us, we're right in the middle of a revolution that IS changing the world right now.
2) We can't keep running AI on von Neuman hardware (nVidia, Cerebras e.t.c.), in the future, lest we want to end up consuming a great part of the worlds electricity solely for training AI models. We need a fundamental hardware change and it's never been more urgent. It's not just an environmental thing, it's also an economical thing. Only insane companies would run their models on nVidia hardware if they could run it on neuromorphic hardware and save heaps of money in investments and also save heaps of money on power consumption.
3) Neuromorphic computing is the best solution to point 2 that we have. I'm not saying that Neuromorphic chips will replace von Neuman architecture in all AI, but where it's feasible it will.
4) Neuromorphic computing is more than just energy savings and economic savings, it's new things that we can do, that we couldn't do before. We're opening up entirely new markets, that only neuromorphic computing is suited for.
5) It's crazy to send all the data from our devices to the cloud to have it analyzed there and having sent back the results to our devices, clogging up the veins of the Internet and delaying the respons, when we can do it on the device. Who can't see this? Neuromorphic computing can really bring some of todays data center capabilities to the device.
6) There are only two really promising Neuromorphic companies on the market and that's Brainchip and GrAI Matter Labs. We're in the middle of a startup dream situation, we only have one real competitor in our space. Everybody would love to have no competition at all, but a situation with one competitor is also really good.
7) We're dead center in the Industrial Revolution 4.0, nothing is more central than AI, AI pervades all the themes of Industrial Revolution 4.0. Who still can't see this? I saw this back in the middle of 2020 when reading "Symbiotic Autonomous Systems White Paper III" by IEEE, AI is in almost every trend they mention. Further as explained in the points above, neuromorphic computing is destined to play a continuously larger role in AI.
Last edited:
Bombersfan
Regular
“Today we have Sean Hehir, CEO of Brainchip, the company that is bringing neuromorphic technology to the entire silicon industry”.
That’s a rather significant call Zach….
D
Deleted member 118
Guest
Sony World Photography Award 2023: Winner refuses award after revealing AI creation
Boris Eldagsen said he used the image to test the competition and to create an "open discussion".
www.bbc.com
perceptron
Regular
Great interview, however, it cuts out at 9.36 minutes. Hoping edge impulse post the second half.
Bombersfan
Regular
Thanks for posting that tech girl. I found that the best thing I’ve seen here in a long time. Listening to Sean talk about ecosystem, and his experience with it really reaffirms to me how they are building this to have as far a reach as possible.
While I would like some individual licenses to to have been signed to help prop the SP a little more, this huge pool of companies all experimenting and complimenting each other’s technology, and the talk of using different foundries and silicon prove outs to guarantee its versatility suggests to me that once everyone figures out how they can all best compliment each other and in which form brainchip will be everywhere. I hope the 1500 and Gen 2 prove outs over the next 4-5 months is what gets the snowball rolling. But who knows it might be longer.
Quercuskid
Regular
I am happy now, I would be happy at $1.50 or $3.00 but would still consider it totally under valued.
Tothemoon24
Top 20
Attachments
FrederikSchack
Regular
Neuromorphic computing is the cutting edge of silicon IP, and @BrainChip_inc is leading the field with its processors. In this episode of ML on the Edge, @zach_shelby sits down with BrainChip CEO Sean Hehir to discuss the tech and its implications for AI: youtu.be/plFQw_G8S68
Attachments
Dhm
Regular
I like part of the introduction which I underlined. Basic for us, but very powerful for those that doubt.
supersonic001
Regular
Sign Up | LinkedIn
500 million+ members | Manage your professional identity. Build and engage with your professional network. Access knowledge, insights and opportunities.
The Pope
Regular
Any reason why this new interview finishes mid sentence at 9min 36sec? Needed some edits due to what was said after this point?
Edit at 1249pm today- the interview is now over 20mins and appears to flow after 9min 36sec
Last edited:
Similar threads
