Hi, how is the train/test split distribution?
* Train: 1457 frames
* Public test: ??? frames
* Private test: ??? frames
Is the final leaderboard scoring only from the private tests?
Thank you
Posted by: rasyidridha @ Feb. 6, 2023, 12:01 p.m.Hi,
We do not disclose the exact number of images in the test sets. However, evaluation runtime limit for the public test set and private test set is 2 seconds per image (1280X1280 pixels) times the number of images, longer runtime will cause a submission error. The instance is used for running and scoring your inference code is g4dn.xlarge (https://aws.amazon.com/ec2/instance-types/g4/).
And indeed, the final scoring results are the performance of up to 2 selected models on the private test set.
Neta,
MAFAT Challenge Team