Hi! Could you please clirify rules about hand labeling, is it allowed? Now media shown on the leaderbord shows submission error from 75 to 175 inline (from 0-370), this media coud be used to correct errors manualy and be submitted or even train models on manuly labeled images what can gives huge score improvement. Would you check top submissions?
TLDR: Is test part hand labeling forbidden and will be checked?
I have a slightly different but related question. Round two is based on test set 2.
Now we have test set 1. are we allowed to hand label test set 1 (since it is not used in the evaluation of round 2)?
2 Likes
Hi,
Re: Hand labelling test set 1 since it is not included in the evaluation of round
This is ok - hand labeling of the test set 1 is acceptable as a way of producing more training data for the machine-learning algorithm to work with.
Re: Using the error maps on the leaderboard to correct errors in the test set, by hand
If this means going in by hand to correct the output of the machine-learning algorithm using the error maps before submitting a final answer for test set 2, it would not be acceptable. It would go beyond the scope of the challenge to have a fully automated ML interpretation.
1 Like