Informations regarding possible submission to Track 3

[Lorenzo Brigato]

Dear organizers,

I am an author of the recently published paper Tune It or Don't Use It: Benchmarking Data-Efficient Image Classification at one of the ICCV Workshops concerning data-efficient image classification.

Our paper proposed a benchmark of six datasets to evaluate methods that were originally proposed to be effective in scenarios where few-labeled examples are available. Indeed, all datasets have less than 100 samples per class.

Among them, we also used a sub-sampled version of the ImageNet training set comprising only 50 images per class. This variant is definitely challenging for modern image classifiers (max accuracy 46.36%). Therefore, we are inclined to believe that our work could bring a useful discussion around the topic of training Imagenet with few samples, something that is hardly experimented with. However, before submitting our paper to Track 3 as it is, I was wondering if our article is anyway eligible despite having experiments that cover other datasets along with ImageNet.

Best regards,

Lorenzo Brigato