Fossil-Eyes

Fossil data output
Getting user input through a video feed

Inspiration

Did you know that most species these days won't leave behind a fossil trace? Our aim is to help scientists with their efforts in documenting what species are endangered, or close to being extinct by looking at fossil samples from the past.

What it does

It identifies what species the fossil is, based on live input from a video feed or an image uploaded onto the website. It will then output some interesting facts about the fossil and display them on the webpage, as well as update a map showing where the fossil was found.

How we built it

HTML, CSS, JavaScript, Python (Flask, openai, PyTorch, Pickle, Scikit-learn). We also experimented with using PHP and the OpenCv library (though we didn't end up using them in the end). We used the openai API in order to identify the fossil that was being inputted and to output some information about the fossil.

Challenges we ran into

We had 6 ideas that we experimented with and tried to make, but ultimately didn't follow through with them as they each had their own set of challenges that we'd require more than the allocated time to overcome. For example, even though we had an key for the OpenAI API, Dall-E proved to be quite challenging to work with, so we scrapped our ideas of replacing our clothes with medieval clothes and replacing furniture with medieval furniture. The openai API we were using also didn't work at certain points in the challenge, making it quite difficult to test at times.

Accomplishments that we're proud of

Actually training and using our own AI model using PyTorch alongside the openai API, as well as starting Fossil-Eyes 9 hours into the hackathon and still managing to finish it in time, especially with us being so sleepy! We also got to learn so much about how computer vision and machine learning is implemented in python.

What we learned

Since this was our first time actually working with AI tools, we were exposed to the various ways it could be implemented by just using some Python. This was quite eye-opening and we learnt a lot about machine learning, computer vision and AI.

What's next for Fossil-Eyes

We aim to expand the scope of the project by adding social media-like features to it - likes, shares, etc. This way people would be able to discover fossils that they may not have seen before.