Inspiration
I was inspired by the growing problem that the salmon population in the Puget Sound Area face. This summer, I visited a fish ladder in Seattle WA, and I was shocked to learn about the extreme challenges that the salmon population face including rising sea temperatures, water pollution, changes in pH levels, and overfishing. The Chinook Salmon in particular is a type of salmon that is listed as endangered in Washington State. To help address this issue in my local community, I created a prototype AI model that predicts the Chinook Salmon population, based on the water temperature. This data can be useful for scientists and researchers, to track the progress they are making in recovering the Chinook Salmon population.
What it does
My prototype AI model predicts the Chinook Salmon population based on the given inputs: summer water temperature and the winter water temperature. After this, the AI model displays a graph showing the predicted Chinook Salmon population verses the actual Chinook Salmon population.
How we built it
I built my prototype AI model on Google Colab. I programmed my neural network using python and the pytorch library. I also created my own training data by creating a csv file and using past year data from websites: https://www.seatemperature.org/north-america/united-states/puget-sound.htm + https://seatemperatures.net/north-america/united-states/puget-sound/ for the average summer and winter water temperatures in the Puget Sound and https://stateofsalmon.wa.gov/statewide-data/salmon/dashboard/ for the Chinook Salmon population in the Puget Sound. I’ve used eleven years of training data to train the AI model. The first step I took for creating my neural network was standardization of the training data so that the neural network could train better. Next, I defined the inputs and outputs for the neural network. After this, I defined the pytorch neural network. Next, I used a criterion and an optimizer to track the losses and improve the AI model. Lastly, I used a graph visualization tool to print the predicted Chinook Salmon population vs. the actual Chinook Salmon population.
Challenges we ran into
When creating my training data, I couldn’t find enough data containing the average water temperatures of the Puget Sound Area. As a result of this, I had less training data to train my AI model. Additionally, there were issues with the formatting of the data set. To fix this problem, I looked at other YouTube videos on how I could fix the issue.
Accomplishments that we're proud of
This was my first time creating an AI model and using the pytorch library, so I am proud to have a working AI model. I am also proud that I was able to debug and fix all the errors that I came across in my code.
What we learned
I learned that learning how to create an AI model is not an easy task and that it takes lots of time. I could have started learning how to create an AI model earlier to reduce the number of problems that I ran into. I also wish that I had made my AI model have the ability to calculate the future population of the Chinook Salmon.
What's next for The Future of the Chinook Salmon
I want to add more training data for the AI model to improve the accuracy of the predictions. I also want to give the AI more input data such as the pollution levels and pH values of the Puget Sound, because there are a lot more factors that go into predicting the Chinook Salmon population other than water temperature. Additionally, I want to have the AI model be able to predict future populations of the Chinook Salmon. Lastly, I also want to apply this AI model to other types of Salmon in the Puget Sound Area.
Log in or sign up for Devpost to join the conversation.