PSS: Progressive Sample Selection for Open-World Visual Representation Learning.

We propose a practical open-world representation learning setting where the objective is to learn the representations for unseen categories without prior knowledge or access to images associated with these novel categories during training. Existing open-world representation learning methods make assumptions, which are often violated in practice and thus fail to generalize to the proposed setting. We propose a novel progressive approach which does not depend on such assumptions. At each iteration our approach selects unlabeled samples that attain a high homogeneity while belonging to classes that are distant to the current set of known classes in the feature space. Then we use the high-quality pseudo-labels generated via clustering over these selected samples to improve the feature generalization iteratively. Experiments demonstrate that the proposed method consistently outperforms state-of-the-art open-world semi-supervised learning methods and novel class discovery methods over nature species image retrieval and face verification benchmarks. Our training and inference code are released. (https://github.com/dmlc/dgl/tree/master/examples/pytorch/hilander/PSS).