Hardware Podcasts

296: Train Me Later

Embedded.fm · 25 Jul 2019 · 45m · Episode 296

AI in practice: one real use of AI described

What this episode covers

Shruthi Jaganathan, a high school intern at NVIDIA, describes building the Green Machine, a computer vision system using a Jetson TX2 and later a Jetson Nano to sort cafeteria waste. She explains how her team trained a neural network on 5,000 labeled images of trash, using TensorFlow and SSD MobileNet to classify items as recyclable, compostable, or trash. The system uses a camera and projector to display color-coded boxes on items, and Jaganathan details the challenges of calibration, data augmentation, and porting the software to the smaller Jetson Nano using JetCam.

AI in practice

Where this conversation describes AI being used to build or deliver a product. Part of the AI in practice tracker.

Engineering & simulationTensorFlow (data augmentation)5000 images

The team used TensorFlow's data augmentation features and custom scripts to generate varied training images, which helped the neural network distinguish between similar items like compostable and reusable utensils despite visual similarities.

“So we wrote a script to add like different backgrounds and like different colors near the objects. So, so that kind of helped with the whole the colored boxes from the projector issue. Um, we did definitely the rotating and the flipping stuff as well.”

Shruthi Jaganathan · Hear it at 29:37

Most relevant to

early-stage companies

Topics

Mentioned

Companies
NVIDIA
Products
Jetson TX2, Jetson Nano
Tools
TensorFlow, SSD MobileNet, JetCam
Also
FIRST Robotics Competition

Related episodes from other shows

Other shows on the same ground, closest first.

  • OnTrack · 47m

    AI Vision Systems in Manufacturing with Deepview's Eliyahu Davis

    Eliyahu Davis, lead engineer at Deepview AI, explains how his company uses deep learning to automate factory inspection, reducing setup time from 40 hours to minutes. Davis details his transition from aerospace software engineering to hardware, describing how he learned Altium Designer through online resources and mentorship to design the camera's PCB. He discusses the technical challenges of running neural network training on edge devices and the future of foundation models in industrial automation.

    Embedded firmwareEDA & CAD tools
    7 chapters

  • The Next Byte Podcast · 26m

    44. How Computer Vision Is Changing Your World

    Hosts Daniel and Farbod discuss how computer vision AI is applied in manufacturing and agriculture, featuring insights from Plain Sight AI and a project by the Technical University of Denmark. They explain Plain Sight AI's three levels of vision systems for detecting hazardous objects on production lines, emphasizing that Level 3 instance segmentation offers superior sorting and counting capabilities. The hosts also detail a robot developed by PhD student Ronja Goldring to autonomously remove weeds from organic dairy farms using laser or electrocution, highlighting the importance of large, diverse training datasets and collaboration with end-users.

    AI & machine learningIndustrial & consumer product design
    5 chapters

  • The Electromaker Show · 20m

    Vintage Tractor goes EV, Beautiful 1978 Radio Upcycle, $10 EdgeAI Camera, and more!

    The host discusses several maker projects, starting with Liam Tronics' full EV conversion of a Massey Ferguson 65C tractor, which uses an 8051 microcontroller and custom PCB for software-controlled clutching. The episode then covers Embedded World 2025's Startup City, highlighting companies like Tactun, Pineberry Pi, and Doodooc. Finally, the host details two upcycling projects: Mukesh Sankher's $10 EdgeAI camera using an ESP32 Fire Beetle 2 and EdgeImpulse for data collection, and Tom Granger's 1978 Sony Fx300 'Jackal' rebuild, which integrates a Teensy, ESP32, and NFC reading capabilities into the vintage chassis.

    Prototyping & iterationEmbedded firmware
    1 chapters

  • The Electromaker Show · 28m

    ML on Arduino, Framework's Open Hardware, Miniature Billiards Game, and More!

    Ian from Electromaker reviews a variety of maker projects, including Valima’s 3D-printed 'Minority Report' kitchen timer and a miniature billiards game by Electro-Japanese. He highlights TinyML applications, such as running heart arrhythmia detection on an Arduino Mega using the Newton platform, and discusses Framework’s initiative to provide old laptop mainboards to developers to foster reuse and reduce e-waste. The episode also covers new event-based vision sensors from Metavision and crowdfunding campaigns for SMD component counters and USB-to-TTL adapters.

    Prototyping & iterationAI & machine learning
    1 chapters

  • The Electromaker Show · 33m

    Pi Zero Handheld, Arduino TensorFlow Weather Prediction, MangoPI RISC-V and More!

    Host Ian reviews a wide range of maker hardware, highlighting a detailed Arduino Nano BLE 33 weather station that uses TensorFlow Lite for air quality prediction, and a DIY RGB mood lamp called Loopy built around the Digispark ATtiny85. He covers the Penkesu, a retro handheld PC powered by a Raspberry Pi Zero 2 W, and the Roundy Pi, a Kickstarter project featuring a circular screen with RP2040 or ESP8266 options. The episode also discusses the MangoPi Nezha MQ RISC-V development board, Aspinity's analog ML chip, and a bio-hybrid fish made from human heart cells.

    Microcontrollers & dev platformsAI & machine learning
    1 chapters

The summary, chapters and topics are generated from the episode’s audio by this directory, not written by Embedded.fm. The episode itself belongs to its publisher.