TreeSavior

• 

  Google Colab

Inspiration

We saw the Ash Trees around the Mac area being affictied with a mysterious pestilence. We decided to take action and built a deep learning model that would determine if a tree had the emerald ash borer.

What it does

It allows you to take a picture from your phone, import it into the deep learning software and it will determine if the tree is diseased or not. In the future, were are planning on implementing it onto mobile devices so that you can take a picture from and app and have it tell you right then if the tree was diseased or not.

How we built it

We built the webscraper we used to gather the images using selenium. We scraped the images off google and saved them into a file which we later uploaded into google drive. We mainly used google colab to train the deep learning model.

Challenges we ran into

Some of the challenges we ran into was the limited knowlege that we had at the beginning of this project. All of us had very limited knowlge of the software we used to day. We had to overcome this barrier by persevering through all the unexpected bugs and errors.

Accomplishments that we're proud of

We accomplished this weekend. We simultaniously built friendships while also creating a pretty darn good deep learning code from almost no knowlege. We also had to overcome the barrier of collecting the data for our machine learning algorithm. Before we settled on selenium, we tried scraping images using Scrapy and Beautiful Soup. We persevered through the challenges and sucessfully collected the images off good.

What we learned

One lesson we learned is that there are always multiple good solutions to every problem. We struggled with the creation of the neural network and the debugging involved. Sometimes we hit dead ends and had to go back, persevere, and attack the problem from a different angle. One example of this is how we imported the images into the deep learning algorithm.

What's next for TreeSavior

We plan to expand the deep learning model by training it on more trees and diseases, not just in ontario. We also plan on deploying the model onto mobile devices through tensorflow lite or tensorflow mobile so that the normal, everyday people can take a picture of a tree on their phone and have it tell them if the tree is diseased or not.

Built With

• googlecolab
• python
• selenium
• tensorflow