Welcome to ACM AI's first Mini-ML Challenge!

In this challenge, you’ll build a model to automatically identify the land use type (e.g. Forest, Highway, Pasture, etc.) from images captured via satellite. The theme around this challenge is *efficiency* since you only have a few hours to train a model.

Your Task

Use the provided train images (with labels) to train your model.

Generate predictions for the test images and submit them in the required format.

Constraints:
- Your model must have fewer than 100 thousand parameters.
- All training and inference must be done in a Kaggle Notebook.
- You must save down the parameter count in your notebook (see starter code).

Your submissions will be scored on a public leaderboard during the competition and a private leaderboard at the end.

All images and CSVs are available under the Data tab. See the Evaluation page for details about the hidden ground truth and leaderboard splits.

Good luck!