Tracking and Mapping Endangered Species

What inspired us

Inspired by the 17 SDG challenges, we chose to address the "Life on Land" goal and decided to create a project that tracks endangered and vulnerable species.

What it does

The project is comprised of three parts: a CNN model that classifies animals, a database storing the times and locations of each animal sighting and a heat map that helps visualise this data.

How we built it

None of us had done any ML before so we used an online tutorial to learn how to create a CNN. We then took this and adapted it for our purposes and conducted tests on 20,000 to assess its accuracy. Alongside this, we created a database of simulated drone data. We then used google maps API to create a heat map of this data in Svelte.

Challenges we ran into

The process of adapting the original model for our purposes took some time and we had to sift through the original code to make it work. It was difficult to improve the accuracy of the model as sometimes we'd try to make improvements that would actually worsen the results.

Accomplishments that we're proud of

We're really proud that we managed to get a working CNN model that works with approximately 85% accuracy. We're also happy that we managed to connect all three parts of the project together: model, database and heat map. We think that we managed to create an intuitive interface for viewing the data.

What we learned

We've learned how to use tools such as github and tensorflow. We had difficulty with syncing files and working together with other people, but in the end learnt how to do this efficiently and make the most of everyone's individual strengths.

What's next for Tracking and Mapping Endangered Animals

We would love to improve our model further, perhaps by using a three block VGG for better recognition. We'd also add classification for more endangered or vulnerable animal types, for example gorillas or rhinos.

Built With

• flask
• github
• javascript
• keras
• python
• sqlalchemy
• svelte
• tensorflow