Exploratory Data Analysis (Training data)

[jessicarose00]

This research project is missing an EDA of the training data, which would improve our understanding the input data and its quality. Ideally this would be done in a jupyter notebook and stored in the notebooks folder. Some potential explorations and visualizations are listed below, but any other ideas related to EDA for image segmentation tasks are welcome.

General data quality

• Inspect the data visually by plotting a few randomly sampled images
• For a single tile, view a histogram for each raw data input (dem, s1 and s2)
• For a single tile, view a histogram for each TML feature

Understanding the data

• Inspect the training data's geographic distribution plotting the locations of the samples (some restrictions apply)
• Understand the label composition / class balance

[jessicarose00]

I've created a notebook dedicated to EDA, so if you decide to work on any of these tasks please add your contributions here: https://github.com/wri/plantation_classifier/blob/model_imprv/notebooks/exploratory_data_analysis.ipynb

[CHRISTINEMUTHEE]

Thanks Jessica. I will have some time on Monday to work on these.

[CHRISTINEMUTHEE]

So i will create a separate branch and work on the specific notebook.