DIY BCI-шлем на Arduino и TinyML: распознаём эмоции силой мысли (ну почти)

Здесь не рассказывают про гигантские суперкомпьютеры и дорогостоящие коммерческие нейроинтерфейсы. В этой статье показан путь от сырой схемы с электродами на лбу до работающего прототипа BCI-шлема, который на Arduino собирает аналоговые сигналы мозга (точнее, лобных долей) и на прошивке TinyML «решает», довольны вы сейчас или испытываете лёгкое раздражение. Всё это — без Biopack, без OpenBCI, с минимумом затрат (пара десятков долларов), но с максимальным погружением в детали: схемы, код, личные промахи и избыточная доза сарказма. Увидев в краудфандинговой рекламе новый «мозговой шлем», автор сначала подумал: «Ну, ещё одна штука для прокрастинации». Но когда узнал, что за $100 можно собрать аналогичную систему самому, захотелось испытать на себе: действительно ли Arduino с несколькими электродами и крошечной моделью TinyML «опознают» эмоцию? Как человек, пробывший инженерный или полубиологический путь (технарь с желанием покопаться в электрическом шуме мозга), автор проверил: да, можно. Хоть и с погрешностями, хотя бы для демонстрации. Впереди — подробная инструкция: какие компоненты взять, как их соединить, куда класть электроды, чтобы не получать случайные сигналы мышечной активности вместо мыслей про кофе, как собрать TinyML-модель, вырезать её под Arduino и запустить «нервный» прогноз вживую. Поехали!

https://habr.com/ru/articles/918016/

#bci #Arduino #tinyml #eeg #эмоции #AD8232 #TensorFlow_Lite_Micro #домашняя_нейронаука

🚀 Why pay more for cloud AI when smarter AI fits in your watch?
Discover how Small Language Models are quietly outperforming LLMs —
• 8X faster
• 90% cheaper
• 100% offline 🤯

From Tesla to smart clinics, this is the AI story no one's telling — yet.
Read the full piece 👇
🔗 https://medium.com/@rogt.x1997/8x-faster-90-cheaper-how-tiny-ai-models-are-beating-gpt-4-at-the-edge-1ae79fe3eb64

#EdgeAI #SLMs #TinyML #FutureReady
https://medium.com/@rogt.x1997/8x-faster-90-cheaper-how-tiny-ai-models-are-beating-gpt-4-at-the-edge-1ae79fe3eb64

Enhancing Remote Sensing with FPGA-Based AI Acceleration

https://rackenzik.com/enhancing-remote-sensing-with-fpga-based-ai-acceleration/

#RemoteSensing #ComputerVision #FPGA #SmallTargetDetection #EdgeAI #SatelliteImaging #Drones #RealTimeProcessing #EnergyEfficientAI #DeepLearning #AIonFPGA #EnvironmentalMonitoring #DisasterResponse #AIEdgeComputing #TinyML #SmartSatellites #AIDrones #GeospatialAI #LowPowerAI #TechForGood

MLP Accelerators are Changing TinyML for Edge Computing

https://rackenzik.com/how-mlp-accelerators-are-changing-tinyml-for-edge-computing/

#TinyML #EdgeAI #MLPAccelerators #FPGAs #EmbeddedAI #AIOnTheEdge #LowPowerAI #MachineLearning #EdgeComputing #AIHardware #AIInnovation #SmartDevices #OnDeviceAI #MLOptimization #TechForGood

Real-time smart tech sparks fresh ideas and bold steps. #Innovation #EdgeAI #TinyML #RealTime #LowPower #SmartTech #Community #FutureTech #TechTalk #SmartDevices
https://medium.com/@sanjay.mohindroo66/edge-ai-tinyml-the-future-of-real-time-smart-devices-5c06424b4271

🌐 Can AI Fit on a Microchip?

TinyML Proves It!

🔥 Why it's revolutionary:
⚡ Real-time processing
🔋 10x energy efficiency
🔒 Data never leaves your device

🌾 Farm sensors → predicts crop health
🏥 Medical wearables → instant diagnostics
🏭 Smart factories → self-repairing machines

Spoiler: It's already here. 👀

Boost to spread the micro-revolution! 🚀

Watch now: https://link.illustris.org/tinyml

#TinyML #EdgeAI #EthicalTech #IoT #LowPowerAI #PrivacyFirst

🤖✨ TinyML: Big AI, Zero Cloud ✨

🌱 Why cloud dependence? #TinyML processes data locally on microcontrollers:

🩺 Privacy-first health monitors
🏭 Predictive maintenance (40% downtime cut)
⚡️ Open-source tools @EdgeImpulse @tensorflow

Build with Arduino Nano/Raspberry Pi! 💻

Full video: https://link.illustris.org/tinyml

#EdgeAI #SustainableTech #EmbeddedSystems

⚡️Boost

OpenAI just announced a new Small Language Model (SLM) --> o3 mini --> let's take a look at current top SLMs and what they can empower with local Vector Databases

Article: https://objectbox.io/top-small-language-models-slms-and-their-power-with-local-vector-databases/

Slide is based on: https://artificialanalysis.ai/downloads/ai-review/2024/Artificial-Analysis-AI-Review-2024-Highlights.pdf

o3 mini announcement: https://www.techmonitor.ai/digital-economy/ai-and-automation/openai-o3-mini-language-model-release?cf-view

#smalllanguagemodel #SLM #LLM #vectordatabase #edgeAI #tinyML

Super excited to share that I’ll be speaking at PyCon+Web 2025 on January 25th in Berlin! 🎤✨ My talk, "𝗘𝗱𝗴𝗲 𝗔𝗜 𝘄𝗶𝘁𝗵 𝗠𝗶𝗰𝗿𝗼𝗣𝘆𝘁𝗵𝗼𝗻 𝗮𝗻𝗱 𝗧𝗲𝗻𝘀𝗼𝗿𝗙𝗹𝗼𝘄 𝗳𝗼𝗿 𝗠𝗶𝗰𝗿𝗼𝗰𝗼𝗻𝘁𝗿𝗼𝗹𝗹𝗲𝗿" is going to dive into the fascinating world where AI meets tiny tech.

It's like fitting an elephant or mammoth into a shoebox! 🐘📦

Can’t wait to meet all the amazing Python and web enthusiasts—let’s make it awesome! 🎉

Details here: https://www.pyconweb.com/activity/edge-ai-with-micropython-and-tensorflow-for-microcontroller

#PyConWeb #PythonFun #EdgeAI #TinyML

I wrote a #rust crate Moden Hopfield Network. I used it to build a neural network that can be trained on the edge. See the demo linked in the README.md.

Modern Hopfield Network has much (much) larger capacity than classical Hopfield Network. They are also called Dense Associative Memory.

#holfield-network #tinyml

https://github.com/dilawar/moden-hopfield-network

Shooting video ahead of my talk at #ParticleSpectra in a couple of weeks time. #IoT #TinyML #AI https://spectra.particle.io/?utm_campaign=spectra_2024&utm_content=webinar&utm_source=x&utm_medium=social

Explore how #TinyML is revolutionizing edge computing by deploying #MachineLearning models on ultra-low-power devices.

Dive into the challenges of optimizing memory, power consumption, and real-time processing in constrained environments: https://ter.li/ndjs0l

Help us out by simply clicking the star ⭐️ button on this repo: https://github.com/harvard-edge/cs249r_book For every 25 stars this open source #TinyML textbook repo gets, @seeedstudio a hardware kit to a school. #Arduino #ESP32 #education #STEM #AI

Huh! Just stumbled across the Analog Devices MAX78000. It's a Cortex M4F with an on-chip CNN accelerator. This looks rather interesting, https://github.com/analogdevicesinc/MaximAI_Documentation. Anyone had a play? How's the development toolchain? #TinyML

#neuromorphic computing and spiking neural networks promise orders of magnitude better energy efficiency than their artificial NN counterparts running on CPUs and GPUs. But we’re still struggling to make SNNs perform as good as ANNs on similar tasks. #AI #TinyML #MachineLearning

Another interesting example of mixture-of-agents architectures, this one from the folks at #EdgeImpulse combining #LLM and #TinyML at the edge. https://www.edgeimpulse.com/blog/using-llms-to-analyze-and-label-satellite-imagery-in-edge-impulse/

By me for @hackster_io, "The CH32, just another 10¢ microcontroller?" The latest project to appear for the 10¢ CH32 RISC-V microcontroller is a tiny wake word engine with impressive accuracy levels. #tinyml https://www.hackster.io/news/the-ch32-just-another-10-microcontroller-a4e5f894ba76

A few days ago I worked out a way of representing neural networks in Rust's type system via const generics and then that lead me to making a neural network library that doesn't use the standard library or any heap allocation, then that lead me to today where I have a pre-trained neural network running on an ATtiny85 approximating the output of an XOR gate, I'm in no way a very good data scientist or embedded engineer at all so the fact I managed to get a neural network running on an 1MHz 8 bit chip with 8k of program memory with 512 BYTES of ram, all whilst doing almost zero optimisations is insane to me

https://github.com/jasonalexander-ja/mynn-attiny-example

#rust #rustlang #tinyml #embedded

By me for @hackster_io, "Benchmarking TensorFlow and TensorFlow Lite on Raspberry Pi 5." The big take away from these new benchmarks is that the Raspberry Pi 5 has similar performance when using TensorFlow Lite to the Coral TPU, displaying essentially the same inferencing speed as Google's accelerator hardware. #ML #TinyML #AI #TensorFlow #RaspberryPi #CoralTPU https://www.hackster.io/news/benchmarking-tensorflow-and-tensorflow-lite-on-raspberry-pi-5-b9156d58a6a2

By me for @hackster_io, "Bringing GPT-4o in From the Cloud to the Edge." #TinyML #SmallAI https://www.hackster.io/news/bringing-gpt-4o-in-from-the-cloud-to-the-edge-289fbf6e0978