How I made an AI powered pocket device
summary
Handheld device uses a Teensy microcontroller for camera and display duties and an ESP32 for wireless connectivity, with a custom PCB and a digital camera module communicating over I2C and a physical button to trigger capture. Images are encoded in a simple C4 format and sent to a Flask REST API where a pre-trained MobileNet V2 runs inference in the cloud. Pipeline yields ~2 second roundtrip and ~90% object-detection accuracy. Teensy + ESP32 split roles MobileNetV2 on Flask REST API
AI-written summary · about
related talks
28mMachine Learning on the Edge
26mWhere can I park? Deep Learning is here to help you!
39mLeveraging the power of C++ for efficient machine learning on embedded devices
21mLeveraging Python for Real-Time Image processing
24mRunning AI/Neural networks on microcontrollers made simple with the STM32Cube.AI
27mHow I've been using JavaScript to automate my house
10mDeep Learning in the Wild: Embedded PyTorch for...
45mBe My Eyes - Leveraging Generative AI to help the blind and visually impaired people