#image-matching-challenge-2024
1 messages · Page 1 of 1 (latest)
@solar lodge This should be fixed now
Thank u. I download the dataset and upload, then solved lol
Hi, I'm running into a strange problem with my submissions. The notebooks run fine and generate the submission.csv with the correct fields and values (compared to a successful earlier submission I've done), but it says Failed: Notebook Threw Exception. When I open the run I can see all cells ran without any exceptions and the output was successfully generated. I looked at the logs and couldn't see any exceptions either and it even says: "Successfully ran in ...". Can someone help me with what might be the issue?
I've had the same problem. I was able to run the imc-2024-submission-example but the DIM-package-submission-example throws errors even though when I run it the notebook runs fine.
Strange, I have no idea why because I checked the two submission.csvs and they are the same format just slightly different numbers. Maybe I should start with copy of the old notebook that works I guess but I'm still interested to know what the reason for that is.
That's a good idea. Did you try both example notebooks?
I haven't tried the example submissions as is, I created one from scrach and the other one was a copy of the imc-2024-submission-example (this one works the one from scratch doesn't)
hmm ok, similar to me then - the imc-2024-submission-example seems to be working fine but nothing else I have tried. I suppose best to start back there then
I guess all submission notebooks must be copied from the example submission, so you will just have to enter again any input you have and copy the code cell by cell. I'm going to try this and update you if it works.
Great - Thank you 🙂 ! I also posted in Discussion so hopefully one of the organizers will be able to answer - I can let you know.
Quick update mine didn't work. It might have to do with the NaN rule that they have to be honest.
Thanks for letting me know. the DIM example notebook is still unsuccessful - will switch back to the light glue example
My best guess, depending on how much time evaluation takes, is that when they rerun it, they do so on CPU, so your code should be CPU compliant.
I've tried to find it but haven't succeeded -- is there a notebook anywhere that has the evaluation code? Like takes my R and t and computes the leaderboard score?
Looks like it's this one: https://www.kaggle.com/code/fabiobellavia/imc2024-3d-metric-evaluation-example
Hello! I started on kaggle with image-matching-challenge today. Is it normal that submission on GPU which took 7 minutes and 8 seconds. After submitting it it runs for 50 minutes already
Yes. I believe the evaluation itself takes a lot of time. It is run on entire test set consisting of scenes of 100's of images (total around 1000 images)
Hi everyone, may I ask how to leverage the training data? Cause I find all notebooks available simply used existing models, not giving a glance at the training data.