We'll probably want to resort to this in the final days to improve our leaderboard position. Require a script to take two (or more) submission csvs and write a new csv that is a weighted average of the originals. This should run faster if we use numpy's loadtxt (skipping the header and the image names and then getting them from settings).
We'll probably want to resort to this in the final days to improve our leaderboard position. Require a script to take two (or more) submission csvs and write a new csv that is a weighted average of the originals. This should run faster if we use numpy's loadtxt (skipping the header and the image names and then getting them from settings).