# train/

This folder contains all the images organized in subfolders representing the class of the image (no of green rectangles in it). The filenames follow the pattern: `<class>_<image_id>.png`.

# train.csv

This file contains a mapping from `image_id` to `count` (the class label). The class label is encoded in three different places (folder-structure, filename, csv file) to enable people to use the DataLoader/Dataset style of their choice.

# test/

This folder contains all the images for which you need to make predictions. The filenames follow the pattern `<image_id>.png`.

# sample_submission.csv

All submissions should follow the format of this file, matching column names (`image_id`, `count`) and row count (121024).