When discussing an idea for a project, we wanted to focus on something that can change the world for the better. We were all passionate about many aspects of life that can be improved, from education to accessibility to medical aid. However, as soon as environment came into the topic, we immediately latched on. After researching on the many environmental problems we could approach we all felt a strong connection to the deforestation problem, not only because it affects a wide range of both places and organisms, but also because the problem is not nearly advertised enough as something that needs to be eliminated.
What it does
Our project is a website that uses machine learning (Google AutoML Vision) to detect whether an image of a forest features deforestation. This website has the ability to grid out a picture into equally-sized, smaller pictures for a more precise measurement of whether deforestation is occurring or not.
How we built it
We used the Google AutoML Vision machine learning platform to train a program to detect whether an image has deforestation or not. In order to do this, we had to collectively take about 270 screenshots on Google Maps Satellite of deforestation and healthy-forest samples to get a large enough dataset for the program to be accurate. In terms of the website development portion, we used Node.js and Bash scripts to send information to the machine learning algorithm, collect it, and display it on the front-end portion of the website.
Challenges we ran into
At first, we didn't think that the sample size for the machine learning program had to have at least 133 elements, so we were very confused when the algorithm predicted images wrong with 100% certainty. We also experienced trouble with sending the information to the machine learning program and collecting it with the Bash scripts; we ended up having to use absolute paths for almost every path mentioned on each script.
Accomplishments that we're proud of
We're very proud of how our project turned out, especially since we have very little experience in machine learning and how it's executed. In addition, actually being able to send data to the machine learning program and being able to collect it was a huge milestone, especially since the method we used (Bash scripts) was not standardly used by others in our situation and caused several authentication errors.
What we learned
We learned a lot about machine learning and how feasible it is to utilize, even as first-year students. We also each learned about different areas of computer science (e.g. web development, scripting) that we weren't particularly experienced with. Additionally, we learned that experienced professionals are more than willing to help and talk to us about what they know, which was a huge help throughout our designing process.
What's next for Project Canopy
In the future, we need to develop it a lot more than our current version, especially since the website doesn't have nearly the amount of features that we want it to have. In addition, we would probably need to improve the machine learning algorithm to detect for the presence of degradation, loss of growth, and more, since those challenges contribute a huge amount to climate change as well.