Animal to Dinosaur Transformer

Inspiration

We noticed how there was significant issue surrounding how dinosaurs were shown to people across the globe. So, since we are all familiar with modern animals, we decided that we could significantly increase the understanding between the relationship between modern animals and dinosaurs, alongside the lifestyle choices that many of the ancient dinosaurs made in the ancient world.

What it does

First, our Animal to Dinosaur Transformer takes in an animal input (e.g. Aardvark, Tiger, Zebra) and outputs the most similar dinosaur counterpart to that animal. Next, we built a machine learning model that predicts features of dinosaurs (e.g. speed, lifespan) based on information of animals alongside their corresponding features.

How we built it

To build our Animal to Dinosaur Transformer, we used the pandas and numpy library to first clean the data, ensuring that it is suitable for usage. Examples include removing any null values in the DataFrames or quantifying certain categorical variables, such as diet. Then, we computed the Euclidean distance between each of the features and returned the name of the dinosaur that has the lowest similarity score, which indicates greater similarity.

As for the machine learning segment, we used PyTorch to implement a regression model to predict dinosaur features based on animal ones.

Challenges we ran into

We struggled the most on implementing the similarity function to output the most similar dinosaur, as we had to figure out how to quantify certain variables, such as the diet of particular animals. Furthermore, we wanted to standardize each of the variables such that one does not provide an extreme amount of influence over another, which we struggled on finding the correct mathematical method in doing so.

Accomplishments that we're proud of

Firstly, we are proud of managing to complete this project within the span of a little over a day. Not only that, we managed to get our first experiences of data cleaning and data preprocessing, alongside creating complex algorithms for this project.

What we learned

We manage to improve our skills in using GitHub and using it as a way to share our workspaces. We learned about multiple different ways to normalize data, such as using a logistic model or a logarithmic function, and data cleaning. Not only that, we learned how to use matplotlib to display and create plots of our results.

What's next for Animal to Dinosaur Transformer

We hope to add more visualizations for this project, especially ones regarding the exploratory data analysis that we performed early on throughout the project.