PADS: Policy-adapted Sampling For Visual Similarity Learning

Abstract

Learning visual similarity requires to learn relations, typically between triplets of images. Albeit triplet approaches being powerful, their computational complexity mostly limits training to only a subset of all possible training triplets. Thus, sampling strategies that decide when to use which training sample during learning are crucial. Currently, the prominent paradigm are fixed or curriculum sampling strategies that are predefined before training starts. However, the problem truly calls for a sampling process that adjusts based on the actual state of the similarity representation during training. We, therefore, employ reinforcement learning and have a teacher network adjust the sampling distribution based on the current state of the learner network, which represents visual similarity. Experiments on benchmark datasets using standard triplet-based losses show that our adaptive sampling strategy significantly outperforms fixed sampling strategies. Moreover, although our adaptive sa

Stats

• citations37
• S2 citations—
• github stars63
• HF likes0
• heat score15.46
• arxiv keyroth2020pads

Related papers

• Effectively Leveraging Attributes For Visual Similarity (2021)3.66
• Deep Unsupervised Similarity Learning Using Partially Ordered Sets (2017)9.76
• Visual Similarity Attention (2019)0.00
• Attributable Visual Similarity Learning (2022)11.68
• Sharing Matters For Generalization In Deep Metric Learning (2020)8.35
• Pairwise Similarity Learning Is Simple (2023)7.50
• Dynamic Sampling For Deep Metric Learning (2020)5.84
• Super-sparse Learning In Similarity Spaces (2017)5.24