The initial learning rate is set to 0.1 for all datasets except ImageNet-1K, and is scheduled to decay by a factor of 10 at the 170th epochs. We also carry out experiments using "hard" and "soft" cosine similarity. For the "hard" cosine similarity, we simply adopt a threshold (0.9 in our experiments) on the score to get binary pseudo labels. While for the "soft" cosine similarity, we directly take the score as soft pseudo labels. The results are presented in table 3.

In this paper, we tackle the problem of novel visual category discovery, i.e., grouping unlabelled images from new classes into different semantic partitions by leveraging a labelled dataset that contains images from other different but relevant categories.

In this paper, we tackle the problem of novel visual category discovery, i.e., grouping unlabelled images from new classes into different semantic partitions by leveraging a labelled dataset that contains images from other different but relevant categories.