BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
Are we still thinking Plumerai is a goer?
Plumerai People Detection for Arm Cortex-A and x86 - Plumerai Docs
docs.plumerai.com
View attachment 9340
View attachment 9338
View attachment 9339
Hi tls,
When Edge Impulse realized they needed Akida:
https://blog.plumerai.com/2021/10/cortex-m-inference-software/
Monday, October 4, 2021
At Plumerai we enable our customers to perform increasingly complex AI tasks on tiny embedded hardware. We’re proud to announce that our inference software for Arm Cortex-M microcontrollers is the fastest and most memory-efficient in the world, for both Binarized Neural Networks and for 8-bit deep learning models. Our inference software is an essential component of our solution, since it directs resource management akin to an operating system. It has 40% lower latency and requires 49% less RAM than TensorFlow Lite for Microcontrollers with Arm’s CMSIS-NN kernels while retaining the same accuracy. It also outperforms any other deep learning inference software for Arm Cortex-M:
| Inference time | RAM usage | |
|---|---|---|
| TensorFlow Lite for Microcontrollers 2.5 (with CMSIS-NN) | 129 ms | 155 KiB |
| Edge Impulse’s EON | 120 ms | 153 KiB |
| MIT’s TinyEngine 1 | 124 ms | 98 KiB |
| STMicroelectronics’ X-CUBE-AI | 103 ms | 109 KiB |
| 77 ms | 80 KiB |
Board: STM32F746G-Discovery at 216 MHz with 320 KiB RAM and 1 MiB flash
Our inference software builds upon TensorFlow Lite for Microcontrollers, such that it supports all of the same operations and more. But since resources are scarce on a microcontroller, we do not rely on TensorFlow or Arm’s kernels for the most performance-critical layers. Instead, for those layer types we developed custom kernel code, optimized for lowest latency and memory usage. This includes optimized code for regular convolutions, depthwise convolutions, fully-connected layers, various pooling layers and more. To become faster than the already heavily optimized Arm Cortex-M specific CMSIS-NN kernels, we had to go deep inside the inner-loops and also rethink the higher-level algorithms. This includes optimizations such as hand-written assembly blocks, improved register usage, pre-processing of weights and input activations and template-based loop-unrolling.
Although these generic per-layer-type optimizations resulted in great speed-ups, we went further and squeezed out every last bit of performance from the Arm Cortex-M microcontroller. To do that, we perform specific optimizations for each layer in a neural network. For instance, rather than only optimizing convolutions in general, our inference software makes specific improvements based on all actual values of layer parameters such as kernel sizes, strides, padding, etc. Since we do not know upfront which neural networks our inference software might run, we make these optimizations together with the compiler. This is achieved by generating code in an automated pre-processing step using the neural network as input. We then guide the compiler to do all the necessary constant propagation, function inlining and loop unrolling to achieve the lowest possible latency.
...
And if the above results are still not fast enough for your application, we go even further. We make our AI tiny and radically more efficient by using Binarized Neural Networks (BNNs) - deep learning models that use only a single bit to encode each weight and activation.
So they are using a software SNN, like MetaTF.
Somewhere they mention 55 frames per second(after all, it is software) - now who can do 1000 fps?
Fact Finder
Top 20
I have just opened a new thread which is not something I have done titled 'BRAINCHIP BENEFACTOR REVEALED'.
I would ask that we keep this new thread entirely about Peter van der Made and his character.
I would suggest that if you are part of any other group on Facebook or the like where appropriate you might also do something similar.
@GrandRhino if I might also suggest that this could be on your AkidaAi site as it is clearly a significant matter relating to this company and its
core values.
And if ASX Investor is out there reading this post perhaps the next time you present on Brainchip you could make mention of this donation by Peter van der Made as the story from the windsurfer is being picked up by other websites the most recent of which was one for those with an interest in defence.
My opinion only DYOR
FF
AKIDA BALLISTA
I would ask that we keep this new thread entirely about Peter van der Made and his character.
I would suggest that if you are part of any other group on Facebook or the like where appropriate you might also do something similar.
@GrandRhino if I might also suggest that this could be on your AkidaAi site as it is clearly a significant matter relating to this company and its
core values.
And if ASX Investor is out there reading this post perhaps the next time you present on Brainchip you could make mention of this donation by Peter van der Made as the story from the windsurfer is being picked up by other websites the most recent of which was one for those with an interest in defence.
My opinion only DYOR
FF
AKIDA BALLISTA
Last edited:
Sartre's daughter said that if her father had belonged to a political party, it would have been one which did not believe it was always right.I just love this line:
“Heavenly guided us through the process of codifying our existential truth”
I would spend my whole day laughing if I worked in marketing.
“OUR EXISTENTIAL TRUTH”
I wonder if the ASX allows existential truth to be published as price sensitive or is it existentially price sensitive.
FF
AKIDA BALLISTA
GrandRhino
Founding Member
Great new thread and added to the timeline. Thanks FF.
Fact Finder
Top 20
Just in case you were wondering this is one persons take on what this all means:
An existential theory of truth
Dale Cannon
Western Oregon State College, USA
Abstract
This article is an attempt to present a simplified account of the theory of truth expressed in the writings of certain existentialist writers - namely, Kierkegaard, Heidegger, Jaspers, and Marcel. It is designed to serve as a supplement to conventional textbook treatments of the nature of truth, which typically ignore the contributions that existentialists have made to the topic. An existential theory of truth stresses the epistemological (not ontological) indeterminateness of meaning and truth, apart from one’s personal participation in determining them. Contrary to superficial interpretations, this theory does not do away either with a transcendent reality or with objectivity. What is rejected is anything that would circumvent the necessary task of participating, oneself, in the epistemological determination of truth.
So now you know.
Regards
FF
AKIDA BALLISTA
Praps it's just as well the local car manufacturing industry folded.Linkedin post put up by The Chief Software Officer at Mercedes Benz - promoting "Hey Mercedes" - and by the looks of it the "Hey Mercedes" function in the EQXX will leap frog the features in the current "Hey Mercedes" function multiple time over plus add to that the power efficiency as well and Smarter
link to the linkedinpost - https://www.linkedin.com/posts/magnus-östberg_dataquality-mercedesbenz-activity-6942423611311861760-d8JG?utm_source=linkedin_share&utm_medium=member_desktop_web
View attachment 9375
Imagine driving along in a Holden and asking "What's that?" while pointing at something.
"Uh! Buggered if I know."
I once got me ontologicals tangled in a quandary - had to hack me way out with a damoclean sword.
Totally agree
Bravo
If ARM was an arm, BRN would be its biceps💪!
Hi @Neuromorphia, great work. Looks brilliant!
If I may, I'd like to suggest adding Nayvya to Speculation High list under Automotive. Reason being that Navya have collaborated with Valeo who provides them with the "brain of the technology" which processes the data collected, detects objects etc. And we all know what that means, don't we?
Violin1
Regular
Traffic in large Asian cities works differently to other places. It works on the principle of fluid dynamics rather than stopping, starting and formal 'giving way'. This is why, when one comes to a major road, one simply steps forward and walks across at a steady pace. The traffic opens and moves around like a liquid flow. However! Stopping or changing pace leads to disaster!
Mea culpa
prəmɪskjuəs
“Fluid dynamics”, the most apt description I’ve read of negotiating Asian city traffic. However, I somehow remain cautious and reluctant to test the process.
Proga
Regular
Bravo,
It looks to me Valeo's third generation Scala lidar incorporates Akida. Won't be commercialised until 2024 which just happens to coincide with Mercedes plans.
Not sure about the new regulations requiring driver and passenger monitoring in 2023 in the EU. I suspect the driver and passenger monitoring systems might be a bolt on in 2023 as a stop gap measure, meaning it won't be incorporated into the current lidar systems. If so, the bolt on driver monitoring system may include Akida in 2023. Fingers crossed.
Last edited:
My understanding was:
1 - if you get your wheel in front of the car beside you, you are entitled to immediate vacent possession, and
2 - if you are blowing your horn when you have a crash, it's not your fault.
Boab
I wish I could paint like Vincent
All you need is one of those very annoying vuvuzela's
Bravo
If ARM was an arm, BRN would be its biceps💪!
Hi Proga,
This chart and info presented within the attached article should help with the implementation time-frames.
One thing that I think is really important to understand in relation to all of this is the the entire electronic control system has to meet funcitonal safety standards ISO 26262, not just the semiconductor devices themselves. I guess this is why we are seeing lots of companies collaborating in order to achieve this goal because the integrated system as a whole must pass muster for compliance to be issued.
Maybe that's a possibility, if safety is better than a human ever can be with the AI, it should recognise the main driver over a 12 year old hijacking mum and dad's car. Also maybe parents can loan the car to their children to drive them to school or a friends etc, setup a geofence where the kids can and cannot go, if the AI becomes smart and we all become passenger then age is pointless.
*singing* Immanuel Kant was a real pissant who was very rarely stable....
If you can't have dreams, you can't have dreams come true!
BaconLover
Founding Member
Reminds me of something I read yesterday.
From rubber to software, Continental epitomises mobility evolution | Automotive World
The CEO of Continental France talks megatrends and 150 years of history with Megan Lampinen
www.automotiveworld.com
Bravo
If ARM was an arm, BRN would be its biceps💪!
I forgot to mention that not all systems are equal and it's clear that those that are inherently more intelligent and capable of "real-time" analysis with a greater ability to integrate with other ADAS systems will TRIUMPH!
Fullmoonfever
Top 20
With Valeo, the one I'm trying to get some time to look into is the Sensor Cortek partnership they have and Corteks partnership with Synopsys.
Like to understand better where they all fit in the lidar production.
Like to understand better where they all fit in the lidar production.
Similar threads
