Interactive Model

Using Neural Networks for Health and Safety

I've been hiking in the hills around San Jose for years, and although it's a fun way to take a break outside of work, I've been warned by friends and family to watch out for mountain lions around the area. I'm pretty sure I don't need an ML model to identify a mountain lion, but I can't say the same for toxic plants!

LeafTracker is an image classification model that differentiates between common non-toxic plants and their toxic look-alikes. The sequential model is built using TensorFlow (Keras) that uses data augmentation and dropout techniques to improve accuracy, and the code for the project can be found on my Github page.

Plant Types

The following lists document the non-toxic and toxic plants covered in the model's dataset.

Non-Toxic Plants

  • Bear Oak
  • Boxelder
  • Fragrant Sumac
  • Jack-in-the-pulpit
  • Virginia Creeper

Toxic Plants

  • Eastern Poison Ivy
  • Eastern Poison Oak
  • Poison Sumac
  • Western Poison Ivy
  • Western Poison Oak

Model Code

Credits

  • The forest photo was taken by Andrew Coelho on Unsplash.
  • Big thank you to Hans Elliott for publishing the Kaggle dataset that made this project possible. More documentation about the dataset can be found on the linked page.
  • TensorFlow's documentation for creating an image classification model has greatly helped me learn how to use the library and complete this personal project.