Make your IoT even Smarter with Tensorflow Lite to Design the Future of Vertical Farming

on YouTube ↗31m

" summary

Demonstrates converting a tiny Keras regression for lettuce weight into a TensorFlow Lite interpreter model, shrinking an 86 KB Keras file to ~20 KB and then an 8 KB quantized C array, and running inference in ~47 ms on a Raspberry Pi motherboard attached to sensors and a camera. Shows feature set (CO2, pH, light, EC, irrigation type), training/normalization steps on laptop, deployment commands, and benchmarks (MobileNet size cut ~65%, latency down ~40%).

AI-written summary · about

" chapters (15)

1. 0:00Start
2. 1:37Outline & What You Will Learn
3. 2:27Indoor Vertical Farming
4. 4:13Irrigation System
5. 5:02System Architecture
6. 5:34Computer Vision
7. 6:19Why Machine Learning?
8. 7:11Why Edge Computing?
9. 12:12ML Workflow with Tensorflow Lite
10. 13:33Setup on your Laptop
11. 14:18Setup on your RPI
12. 25:10Tensorflow Lite Benchmark
13. 25:52Tensorflow Lite Limitations
14. 26:47TFLite on Microcontrollers
15. 29:21Summary

" related talks

• 27mMachine Learning on the edge using TensorFlow LiteHåkan Silfvernagel · DevFest WI 2020
• 34mTensorFlow LiteNatalie Pistunovich · Mobile Era 2018
• 17mBuild a smart IoT device with TensorFlow Lite and Raspberry PiKhanh LeViet · Google I/O 2022
• 01h 01mML and the IoT: Living on the EdgeBrandon D Satrom · NDC London 2020
• 54mSmarter Everything with ML & the IoTBrandon D Satrom · NDC Oslo 2022
• 55mUsing TensorFlow for Performing Deep Learning at the EdgeAditya Bhattacharya · Developers Summit 2021
• 59mML and the IoT: Living on the EdgeBrandon D Satrom · NDC Oslo 2019
• 32mAI for Mobile and IoT Devices: TensorFlow LiteSarah Sirajuddin, Tim Davis · Google I/O 2019