Are Labels Needed for Incremental Instance Learning?

In this paper, we learn to classify visual object instances, incrementally and via self-supervision (self-incremental). Our learner observes a single instance at a time, which is then discarded from the dataset. Incremental instance learning is challenging, since longer learning sessions exacerbate forgetfulness, and labeling instances is cumbersome. We overcome these challenges via three contributions: i). We propose VINIL, a self-incremental learner that can learn object instances sequentially, ii). We equip VINIL with self-supervision to by-pass the need for instance la-belling, iii). We compare VINIL to label-supervised variants on two large-scale benchmarks [6], [32], and show that VINIL significantly improves accuracy while reducing forgetfulness.