Added functionality for validation split

[Tarandeep97]

Closes #654: Implement Inria Dataset Validation Set Split

Summary:
This pull request addresses issue #654, which involves implementing the validation set split for the Inria dataset as specified in the contest guidelines (https://project.inria.fr/aerialimagelabeling/contest/).

Changes Made:
Added functionality to the inria.py file.

Details:
Modified the inria.py file to implement the validation set split. The first five images of each location can now be removed by setting split argument to valid.

[Tarandeep97]

@microsoft-github-policy-service agree

[adamjstewart]

Can we call it "val" instead of "valid"? That's what most other datasets call it.

[Tarandeep97]

changed "valid" to "val", as per convention.

[adamjstewart]

Still a lot of failing tests, see https://torchgeo.readthedocs.io/en/stable/user/contributing.html for tips to fix these.

[Tarandeep97]

Hi @adamjstewart, sorry for late reply. I was occupied with my thesis project. I am working on the feedback you have provided.

[adamjstewart]

@Tarandeep97 no worries, I also just finished my thesis. Hope everything went okay!

[Tarandeep97]

@adamjstewart yes everything went well. Thank you. Hope, your thesis also went great.

I have made few changes in the file, as per your feedback. I followed checklist provided here.
I investigated that test cases written fails when we try to fetch files other than austin[1-5].tif. I think this is because of /tests/data/inria/data.py file that generates sample dataset. Even though I tried making changes to /tests/data/inria/data.py, it continues generating same set of images. I tried clearing pytest cache, but it didn't work.
Can you help me guide further, how to fix this problem?

[adamjstewart]

I made the regex a bit simpler and got the tests working, let me know what you think!

[Tarandeep97]

This looks perfect to me @adamjstewart . Thank you for helping!