Inspiration
Our team was heavily inspired by the growing field of computer vision, and feeding AI images in order to fulfill certain tasks.
What it does
Our project builds a model that takes in an image url and reads its pixels in order to classify a dinosaur image as a specific species.
How we built it
We built this model by transforming every single image working within the model to fit a certain template (each image must be 224x224 and in grayscale). We then trained a naive bayes model using over 100 dinosaur images, which can be interacted with through our user interface, which was created using tkinter.
Challenges we ran into
Most of our problems were caused by our lack of experience, causing most of our time to be used researching the things that we had zero experience in, such as using tkinter and working with images in python.
Accomplishments that we're proud of
We're proud that our model works, with relatively high testing and training accuracy! We are also amazed that our user interface successfully takes in user input and responds correspondingly!
What we learned
We learned that we should limit ourselves by our inexperience! Following this project, our members have gained a new confidence to be able to tackle projects beyond our expertise!
What's next for Tim the AI Dino-expert!
If we were to work further with this project, we would train the model using much more complex features, rather than just reading the direct pixels of the image. Additionally we would customize and beautify our user interface to be much more lively!
Log in or sign up for Devpost to join the conversation.