Mega-Resolution

• 

  Patates Belges à l'AI

Inspiration

Being students from Polytechnique Montréal we are proud to invest our time into sustainable development. This project was about meteorological conditions, which was a new challenge for us. We used some knowledge we had on meteorology to improve our understanding.

What it does

The visualization part loads the data, then flattens and normalizes each array of the features. The vectors are put in a dataframe, which will be used to make a correlation matrix. The model takes in input some climate information and predicts the temperature change.

How we built it

We used Facets to get some graph on the data. We used GAN to create a benchmark linear. We used SRGAN to create a solution specialized for our problem. We used LSTM to keep the information in time.

Challenges we ran into

For the visualization part, we ran into some problems with correlation management. For the machine learning part, we ran into some space and speed issues because of the large size of the data and the high resolution of the images.

Accomplishments that we're proud of

Create some full model with low experience team (4 dev).

What we learned

Facets, DataVisualization, Machine learning, constitutional network and LSTM

What's next for Mega-Resolution

Our training process speed would benefit from using a AWS pipeline rather than using google colab. The time consumed can also be reduced with data reduction. We can improve the final model prediction by adding pre-trained model and other meteorological variables.

Built With

• colab
• facets
• matplotlib
• pandas
• python