We are given Extra images data and the size is a little big. Therefore, we should use this data for better score.
Then, I have a question. How can we use new turtle ids("new" means not in tarin.csv) in Extra images data ?
If Ids in Extra data are also in original train data, we can merge these ids to original train.csv.But some ids are not in original train data.
If you have any idea, please teach me.
Hello,
You can utlize constrastive learning to learn structural representations, such that the model parameters can be finetuned on a downstream classifier.
https://arxiv.org/abs/2111.14271
https://arxiv.org/abs/2004.11362