Sparse annotations for training

[SebastienTs]

Is there currently a way to use sparse annotations for training, i.e. for pixel classification by ignoring all the pixels set to 0 in the target image during loss computation and only considering pixels set to a value >0? If that is not possible, what would be the minimal modification to the code to achieve this (at least for some, or ideally for all existing losses)?

[mdraw]

Hello @SebastienTs, yes this is possible with the regular torch.nn.CrossEntropyLoss by setting the ignore_index parameter to the label ID that you want to ignore. Since 0 is often reserved for a background class I usually use other ID values. It just has to be the ID that is used in the label files; no changes to the model outputs are needed.

The DiceLoss in elektronn3 doesn't support an ignore_index option but you can use the weight parameter to set the channel weight of the class that is to be ignored to 0.

[SebastienTs]

Great! Thanks so much Martin!

…

On Thu 17 Aug 2023, 21:09 Martin Drawitsch, ***@***.***> wrote: Hello @SebastienTs <https://github.com/SebastienTs>, yes this is possible with the regular torch.nn.CrossEntropyLoss by setting the ignore_index parameter to the label ID that you want to ignore. Since 0 is often reserved for a background class I usually use other ID values. It just has to be the ID that is used in the label files; no changes to the model outputs are needed. The DiceLoss in elektronn3 doesn't support an ignore_idx option but you can use the weight parameter to set the channel weight of the class that is to be ignored to 0. — Reply to this email directly, view it on GitHub <#51 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABTFCUG7D2DZ6B4KC66BBBDXVZT5HANCNFSM6AAAAAA3UPGTAE> . You are receiving this because you were mentioned.Message ID: ***@***.***>