Inspiration ✨

Whether you want to go for a walk, organize a pet race in your backyard, or know what to wear for the day, taking the weather into consideration is crucial. Although sometimes it’s easy to guess what the weather will be, this unfortunately isn't always the case. We found it cool that people could get accurate weather predictions by using their phones to scan the clouds above their heads. Here is why we believe our project is awesome:

1️⃣ It’s a more fun way to get a weather forecast.

2️⃣ We don’t overload the user with unnecessary and complex information such as humidity, wind, or precipitation. We give them only what they need: the state of the weather in the immediate future (rainy, fair, warm, stormy, etc.)

3️⃣ Weather predictions are often made on the scale of cities or wider areas. However, the weather is not always the same throughout a city, and it may even be different from one neighborhood to another. Cloudnerd solves this problem by giving a prediction based on the sky above your head, thereby providing more accurate location-based forecasts.

What it does 🐉

Our web application is quite straightforward. Once you open it, you can select the camera of your choice through the settings icon. Then scan the clouds near you and the model will give you a weather forecast. Et voilà!

How we built it ⚙️

We built the web app with React by using Typescript. For the machine learning model, we made it with Tensorflow. Our approach was to train the model to recognize the different types of clouds. Since there are 10 major types of clouds on Earth and each of them gives us an accurate prediction of the weather in the future, our approach was to build a machine learning model capable of distinguishing between these types of clouds in order to get the corresponding weather forecast. Our accuracy rate, for now, is 84%. The web app was deployed online using Firebase.

Challenges we ran into 🔥

One of our biggest problems when dealing with this project was to conciliate the efficiency of the model with the user experience. As you may have seen, the web app can take some time to load and is still a bit resource intensive. We have done our best to make the user experience smoother, but we are aware that there are still some flaws that need to be fixed.

Accomplishments that we're proud of 😁

We managed to build a web application with an enjoyable user interface and a fairly efficient model. Our approach to solving weather forecasting problems is original and we feel that our execution has been good.

What we learned 📖

We learned a lot about clouds. The different types of clouds and what they tell about future weather. We also learned how to better prepare and clean data before building the machine learning model.

Built With

• firebase
• react
• tensorflow
• typescript