Details: this project involves segmenting an image to distinguish crop rows. Whether we use a computer vision approach or a deep learning approach (U-net) is up to the team and the data available. This model will be built from scratch – expect a lot of work with tensor shapes, reading articles, running small experiments.
Team size: 3
Progression:
- gathering dataset
- creating a U-net for a commonly used dataset (like ImageNet)
- cleaning the data
- making U-net work on crop row dataset
- tuning hyperparameters
Technologies: Pytorch, Python, (maybe a little bit of W&B if we have time at the end)