You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Something-something-v2 video dataset is splitted into 3 meta-sets, namely, meta-training, meta-validation, meta-test. Overall, dataset includes 100 classes that are divided according to CMU [1] The code also provides a dataloader in order to create episodes considering given n-way k-shot learning task. Videos are converted to the frames under sp…
This repository contains the main ResNet backbone experiments conducted in the ICLR 2022 spotlight paper "On the Importance of Firth Bias Reduction in Few-Shot Classification".
This repository contains the firth bias reduction experiments on the few-shot distribution calibration method conducted in the ICLR 2022 spotlight paper "On the Importance of Firth Bias Reduction in Few-Shot Classification".
This repository contains the experiments conducted in the ICLR 2022 spotlight paper "On the Importance of Firth Bias Reduction in Few-Shot Classification".
[CVPR-2022] ''Doodle It Yourself: Class Incremental Learning by Drawing a Few Sketches'', IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), 2022.