Mining on Manifolds: Metric Learning without Labels
This addresses the challenge of reducing annotation costs for fine-grained visual tasks by providing an unsupervised method that can enhance model performance without labeled data.
The paper tackles the problem of unsupervised hard training example mining by identifying positive and negative examples based on disagreements between Euclidean and manifold similarities, enabling fine-tuning of pre-trained networks for tasks like fine-grained classification and object retrieval, with results showing performance on par or better than prior supervised models.
In this work we present a novel unsupervised framework for hard training example mining. The only input to the method is a collection of images relevant to the target application and a meaningful initial representation, provided e.g. by pre-trained CNN. Positive examples are distant points on a single manifold, while negative examples are nearby points on different manifolds. Both types of examples are revealed by disagreements between Euclidean and manifold similarities. The discovered examples can be used in training with any discriminative loss. The method is applied to unsupervised fine-tuning of pre-trained networks for fine-grained classification and particular object retrieval. Our models are on par or are outperforming prior models that are fully or partially supervised.