about results in IMC2022

[KN-Zhang]

Hi, I’m conducting experiments on the IMC2022 competition in Kaggle. However, the final metric I received is the 'score', which represents the average accuracy over a range of thresholds (as shown in the figure). How can I obtain the mAA@10 value as reported in Table 3?

[Parskatt]

The notebook is public since a while back, you can see the approach there.https://www.kaggle.com/code/johanedstedt/tiny-roma-imc2022?scriptVersionId=130698232

[KN-Zhang]

Yes, I’ve reviewed the notebook, but I can only generate a submission.csv file that includes three sample image pairs. The results of full test set of 10,000 image pairs from the hidden test set is not accessible.

[Parskatt]

If you are able to generate output for 3 pairs you just change it to not dryrun and submit it no?

[KN-Zhang]

If you are able to generate output for 3 pairs you just change it to not dryrun and submit it no?

Thanks, I am runing it. But I just want to ensure if the output score the same metric as mAA@10 in Table 3? I see the public score of the notebook you mention is 0.88034, which is the same as mAA@10 in Table 3. :)

[Parskatt]

If you are able to generate output for 3 pairs you just change it to not dryrun and submit it no?

Thanks, I am runing it. But I just want to ensure if the output score the same metric as mAA@10 in Table 3? I see the public score of the notebook you mention is 0.88034, which is the same as mAA@10 in Table 3. :)

Yes, that's the same metric. There is more info on how it's computed on kaggle :)