Deep Unsupervised Similarity Learning using Partially Ordered Sets
This addresses the challenge of fine-grained visual similarity learning in computer vision without labeled data, but it is incremental as it builds on existing unsupervised methods.
The paper tackles the problem of unreliable and contradictory pairwise or triplet relations in unsupervised similarity learning by grouping samples into surrogate classes and using local partial orders to link classes, formulating similarity learning as a partial ordering task with self-supervision. The approach shows competitive performance on detailed pose estimation and object classification.
Unsupervised learning of visual similarities is of paramount importance to computer vision, particularly due to lacking training data for fine-grained similarities. Deep learning of similarities is often based on relationships between pairs or triplets of samples. Many of these relations are unreliable and mutually contradicting, implying inconsistencies when trained without supervision information that relates different tuples or triplets to each other. To overcome this problem, we use local estimates of reliable (dis-)similarities to initially group samples into compact surrogate classes and use local partial orders of samples to classes to link classes to each other. Similarity learning is then formulated as a partial ordering task with soft correspondences of all samples to classes. Adopting a strategy of self-supervision, a CNN is trained to optimally represent samples in a mutually consistent manner while updating the classes. The similarity learning and grouping procedure are integrated in a single model and optimized jointly. The proposed unsupervised approach shows competitive performance on detailed pose estimation and object classification.