Sampling Matters in Deep Embedding Learning

Abstract

Deep embeddings answer one simple question: How similar are two images? Learning these embeddings is the bedrock of verification, zero-shot learning, and visual search. The most prominent approaches optimize a deep convolutional network with a suitable loss function, such as contrastive loss or triplet loss. While a rich line of work focuses solely on the loss functions, we show in this paper that selecting training examples plays an equally important role. We propose distance weighted sampling, which selects more informative and stable examples than traditional approaches. In addition, we show that a simple margin based loss is sufficient to outperform all other loss functions. We evaluate our approach on the Stanford Online Products, CAR196, and the CUB200-2011 datasets for image retrieval and clustering, and on the LFW dataset for face verification. Our method achieves state-of-the-art performance on all of them.

Add supplementary material. Paper published in ICCV 2017

References in corpus (10)

• Adam: A Method for Stochastic Optimization
• Deep Residual Learning for Image Recognition
• Learning Face Representation from Scratch
• DeepID3: Face Recognition with Very Deep Neural Networks
• Learning Deep Embeddings with Histogram Loss
• Local Similarity-Aware Deep Feature Embedding
• The Power of Normalization: Faster Evasion of Saddle Points
• Beyond Convexity: Stochastic Quasi-Convex Optimization
• Learning a Metric Embedding for Face Recognition using the Multibatch Method
• Hard-Aware Deeply Cascaded Embedding

Cited by in corpus (10)

• Learning to rank music tracks using triplet loss
• Improving Generalization via Scalable Neighborhood Component Analysis
• PhishGAN: Data Augmentation and Identification of Homoglpyh Attacks
• Learning to embed semantic similarity for joint image-text retrieval
• Distributionally Robust Deep Learning using Hardness Weighted Sampling
• Generalization in Metric Learning: Should the Embedding Layer be the Embedding Layer?
• Deep Metric Learning with Hierarchical Triplet Loss
• MTGAN: Speaker Verification through Multitasking Triplet Generative Adversarial Networks
• Acceleration of Large Margin Metric Learning for Nearest Neighbor Classification Using Triplet Mining and Stratified Sampling
• Learning a Unified Embedding for Visual Search at Pinterest