lightly-ai
Add RandomFrequenceMask transform
Hi @guarin , I'm pretty new to open source and would like to contribute to this. I have been looking at the Contributions.md for basic guidelines, where should I go next on this? Also are you on the hacktober's discord? (If it's comfortable with you to disclose that? If I had some doubts, I thought I could contact you their itself)
Hi @payo101, the next steps are: 1) We'll assign the issue to you; 2) You open a pull request as outlined in CONTRIBUTIONS.md. For more information on the implementation details please check out the issue referenced in the description. For reference you can also take a look at the augmentations we have already implemented (see, e.g. random rotation). Could you please quickly confirm if you're interested in working on this?
Also are you on the hacktober's discord? (If it's comfortable with you to disclose that? If I had some doubts, I thought I could contact you their itself)
Forwarding this to @guarin :)
Also are you on the hacktober's discord? (If it's comfortable with you to disclose that? If I had some doubts, I thought I could contact you their itself)
Yes you can find me under the same name :)
We also have a discord for lightly related questions: https://discord.gg/xvNJW94
Hi @guarin , So I have 2 doubts to ask
- Will the input be the RFFT of the Image, or the Image itself? I have been currently working with assuming the former
- Should the output also return the binary mask that's being applied on the FFT?
Also, I am facing a mypy error while proceeding to commit
After looking into it, all necessary steps that can be done to prevent this has been done, but it's still showing this error, any chance anyone could help me with it?
Also, I am facing a mypy error while proceeding to commit
After looking into it, all necessary steps that can be done to prevent this has been done, but it's still showing this error, any chance anyone could help me with it?
Hey @payo101, it's interesting that this only comes up now. You can fix it by adding an ignore statement:
class RandomResizedCropAndFlip(nn.Module): # type: ignore[misc] # Class cannot subclass "RandomVerticalFlip" (has type "Any")
If you look at the other examples just above, this is exactly how we fixed it previously. Strange that mypy checks on master are passing... Maybe it's also related to mypy version. Can you share which version you're using?
it's 1.4.1
Ok, that's correct. Did you manage to fix the problem with the proposed # type: ignore ...?
Hi @guarin , So I have 2 doubts to ask
- Will the input be the RFFT of the Image, or the Image itself? I have been currently working with assuming the former
- Should the output also return the binary mask that's being applied on the FFT?
Could you also clarify on these two doubts also @philippmwirth Also, will the k% which is mentioned for the transform be a user input, I am working under the assumption that it will be?
You can assume the input is the RFFT tensor. No need to return the binary mask.