2022 IEEE GRSS Data Fusion Contest Track SLM Forum

Go back to competition Back to thread list Post in this thread

> What is allowed regarding training data?

Dear organizers,

In the instructions I haven't found restrictions regarding the training data that we can or cannot use. Are we strictly limited to the training set that is given in this competition, or we can also pretrain our model on other datasets (other than ImageNet) ?

Thank you

Posted by: mapaf2 @ Jan. 16, 2022, 10:14 p.m.

You can use other data / tasks for pretraining as long as this data is available in general (e.g. use of Sentinel data ok, manually labeling not ok). You should not violate the semi-supervised idea (too much), i.e. semantic labels are only available for the defined part of the training set. In any case, you need to be open about it. Before the begin of the test phase you will be required to submit a short description of your approach in which you need to state all additional data sources and how they have been used.

Posted by: rhaensch @ Jan. 24, 2022, 7:10 p.m.
Post in this thread