Mining on Manifolds: Metric Learning Without Labels
By Ahmet Iscen, Giorgos Tolias, Yannis Avrithis, Ondřej Chum. 2018
122 citations recorded by the source corpus.
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.
Subfields: Representation & Self-Supervised Learning
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Representation & Self-Supervised Learning
- Diane Larlus
- Ahmet Iscen
- Shiliang Zhang
- Dongkai Wang
- Yannis Avrithis
- Giorgos Tolias
- Ondřej Chum
- Yannis Kalantidis
Citations in the corpus
- Unsupervised Embedding Learning via Invariant and Spreading Instance Feature (2019, methodology): “MOM [21] is the only method that claims for unsupervised metric learning.”
- Label Propagation for Deep Semi-Supervised Learning (2019, background): “amples [42], relations between examples and cluster centroids [1], or considering the manifold structure of data [19].”
- Dual Temperature Helps Contrastive Learning Without Many Negative Samples: Towards Understanding and Simplifying MoCo (2022, background): “A key property of CL is that it involves negative samples, and a major line works [3,10,26,28,29,36,40,46,49,53,61] have shown that mining hard negative samples can be beneficial for performance.”
- Hard Negative Mixing for Contrastive Learning (2020, background): “[23] mine hard negatives from a large set by focusing on the features that are neighbors with respect to the Euclidean distance, but not when using a manifold distance defined over the nearest neighbor graph.”
- Unsupervised Person Re-Identification via Multi-Label Classification (2020, background): “[11] utilizes manifold learning to seek positive and negative samples to compute the triplet loss.”