Late submission for working notes

#9
by chychiu - opened

Can we submit now against the public test set to help us write the working notes? I am aware it won't be ranked but would be good to get official benchmarked performance for us to discuss in the notes nonetheless

Bohemian Visual Recognition Alliance org

Hi @chychiu ,

The late submissions were allowed just now. We had to downsize from the GPU instance to pure CPU, but we increased the time limit to 2 days.
PS: The number of submissions remained the same.

Best,
Lukas

@picekl I'm getting an error when trying to make a submission: SyntaxError: Unexpected token 'I', "Internal S"... is not valid JSON

I'm pretty sure this is the same message I received when the competition prematurely ended before, so it looks like submissions aren't working?

Bohemian Visual Recognition Alliance org

@picekl I'm getting an error when trying to make a submission: SyntaxError: Unexpected token 'I', "Internal S"... is not valid JSON

I'm pretty sure this is the same message I received when the competition prematurely ended before, so it looks like submissions aren't working?

Thx for reporting it @jack-etheredge . It should work fine at the moment. The competition token was automatically invalidated due to some "suspicious activity".

LP

This comment has been hidden

Will the submissions close Friday night? That seems logical, but I'm a bit concerned due to how long things are taking to evaluate on CPU. I don't know if it makes sense to leave them open a bit longer considering that the camera-ready deadline is much later than the submission deadline. That said, I have no idea what it's costing to make the submissions available.

Edit: I'm going to leave this up in case the answer is important to anyone else, but I think if I run some smaller submissions I can work around things. Some of my ablations and thresholding experiments were using a full ensemble of models with multiple test time augmentations, but if I run an individual model and strip out augmentations that are unnecessary for the specific comparison as a baseline, I can make these go much faster. Lessons learned for next time around.

Bohemian Visual Recognition Alliance org

Hi @jack-etheredge ,

You can leave it running and update the paper once you have the results. However, you must submit the first version on time and make it available for reviewers "on time."
I will send the papers to reviewers on the weekend but I assume some might access it after the weekend so you can try your luck :D

Best,
Lukas

Sign up or log in to comment