Inspiration
Upon encountering the problem, we tried to find some useful academic papers to solve our problem. We thus found a few useful neuron network tools such as mask-rcnn, unet etc. Our learning part is mostly dependant on that.
What it does
Our project realized a frontend visualization and a backend analysis of the geo data provided.
How we built it
Challenges we ran into
- Environment problem. Some of the suggested packages lack maintainence and we thus encountered quite a few challenges. The dataformat is also a bit hard to deal with.
- Badness of raw dataset, some of the dataset are rather a bit chaotic and lack high resolution and thus our work cannot detect them. We haven't really solved this problem, but we may foresee a solution with some more working hours.
Accomplishments that we're proud of
We are satisfied with what we had as a team of merely 2 people.
What we learned
A few new framework, both frontend and backend analysis of compter vision.
What's next for Solar panel potential estimation and visualization on cv.
- Refining the detection procedure.
- Adding some categorization approaches, such as decision tree and naive bayesian to prepross the data set.
- Finding some more and better visualization to express what happens in the real life.
Log in or sign up for Devpost to join the conversation.