Member-only story

Custom Edge TPU SSDMobileNet for Shooting Toy Reindeer with DIY 3D Printed Robot

Steven Smiley
9 min readNov 25, 2022

--

What if your DIY robot could leverage a custom built object detector model to accomplish tasks such as playing fetch or shooting a specific object?

Table of Contents (TOC)

  1. Purpose
  2. Background
  3. Materials & Methods
  4. Results & Conclusions
  5. References

1. Purpose

The purpose of this article is to create a custom object detector tflite model on my toy reindeer, by applying transfer learning. Transfer learning is applied to a state of the art object detector, SSD MobileNet V2 FPNLite.³ The trained model should provide reasonable precision and recall when deployed to a Raspberry Pi 4 Model B with Coral USB Accelerator for Edge TPU inference.¹⁵

Training Notebook

https://nbviewer.jupyter.org/github/stevensmiley1989/MrRobot/blob/main/Jupyter%20Notebooks/augmented_ssd_MobileNet_v2_FPNLite_320x320_5_8_2021_reindeer.ipynb

Evaluation Notebook

--

--

Steven Smiley
Steven Smiley

Written by Steven Smiley

Lead Machine Learning Engineer who also enjoys writing about Data Science, CV, DL, ML, AI, Python https://www.linkedin.com/in/stevensmiley1989/

No responses yet