Generalization in Metric Learning: Should the Embedding Layer Be Embedding Layer?

@article{Vo2018GeneralizationIM,
  title={Generalization in Metric Learning: Should the Embedding Layer Be Embedding Layer?},
  author={Nam S. Vo and James Hays},
  journal={2019 IEEE Winter Conference on Applications of Computer Vision (WACV)},
  year={2018},
  pages={589-598}
}
  • Nam S. VoJames Hays
  • Published 8 March 2018
  • Computer Science
  • 2019 IEEE Winter Conference on Applications of Computer Vision (WACV)
This work studies deep metric learning under small to medium scale as we believe that better generalization could be a contributing factor to the improvement of previous fine-grained image retrieval methods; it should be considered when designing future techniques. In particular, we investigate using other layers in a deep metric learning system (besides the embedding layer) for feature extraction and analyze how well they perform on training data and generalize to testing data. From this study… 

The Group Loss for Deep Metric Learning

Group Loss is proposed, a loss function based on a differentiable label-propagation method that enforces embedding similarity across all samples of a group while promoting, at the same time, low-density regions amongst data points belonging to different groups.

The Group Loss++: A Deeper Look Into Group Loss for Deep Metric Learning

This work proposes Group Loss, a loss function based on a differentiable label-propagation method that enforces embedding similarity across all samples of a group while promoting, at the same time, low-density regions amongst data points belonging to different groups.

Dynamically Modulated Deep Metric Learning for Visual Search

This paper proposes a new soft-binomial deviance loss that helps to capture the feature similarity space at multiple granularities and achieves state-of-the-art performances on standard benchmark datasets.

Deep Metric Learning via Adaptive Learnable Assessment

A sequence-aware learnable assessor is proposed which re-weights each training example to train the metric towards good generalization and demonstrates the effectiveness of the proposed approach on widely used CUB-200-2011, Cars196, and Stanford Online Products datasets.

Semantic Granularity Metric Learning for Visual Search

Learning Embeddings for Product Visual Search with Triplet Loss and Online Sampling

This paper proposes learning an embedding function for content-based image retrieval within the e-commerce domain using the triplet loss and an online sampling method that constructs triplets from within a minibatch that significantly outperforms the state-of-the-art on the DeepFashion dataset.

Visualizing Deep Similarity Networks

The visualization shows how similarity networks that are fine-tuned learn to focus on different features and generalize the approach to embedding networks that use different pooling strategies and provide a simple mechanism to support image similarity searches on objects or sub-regions in the query image.

Adapting Grad-CAM for Embedding Networks

This work proposes an adaptation of the Grad-CAM method for embedding networks, and develops an efficient weight-transfer method to explain decisions for any image without back-propagation.

Visualizing How Embeddings Generalize

In observing many triplet selection strategies for Metric Learning, it is found that the best performance consistently arises from approaches that focus on a few, well selected triplets.

Stochastic Class-Based Hard Example Mining for Deep Metric Learning

The key idea is to adopt class signatures that keep track of feature embedding online with minor additional cost during training, and identify hard negative example candidates using the signatures and achieves the state-of-the-art performance on the several standard benchmark datasets.

References

SHOWING 1-10 OF 40 REFERENCES

Attention-based Ensemble for Deep Metric Learning

An attention-based ensemble, which uses multiple attention masks so that each learner can attend to different parts of the object, which outperforms the state-of-the-art methods by a significant margin on image retrieval tasks.

Sampling Matters in Deep Embedding Learning

This paper proposes distance weighted sampling, which selects more informative and stable examples than traditional approaches, and shows that a simple margin based loss is sufficient to outperform all other loss functions.

BIER — Boosting Independent Embeddings Robustly

The metric learning method improves over state-ofthe- art methods on image retrieval methods on the CUB-200-2011, Cars-196, Stanford Online Products, In-Shop Clothes Retrieval and VehicleID datasets by a significant margin.

Deep Metric Learning via Facility Location

This paper proposes a new metric learning scheme, based on structured prediction, that is aware of the global structure of the embedding space, and which is designed to optimize a clustering quality metric (NMI).

Deep Image Retrieval: Learning Global Representations for Image Search

This work proposes a novel approach for instance-level image retrieval that produces a global and compact fixed-length representation for each image by aggregating many region-wise descriptors by leveraging a ranking framework and projection weights to build the region features.

CNN Image Retrieval Learns from BoW: Unsupervised Fine-Tuning with Hard Examples

This work proposes to fine-tune CNN for image retrieval from a large collection of unordered images in a fully automated manner and shows that both hard positive and hard negative examples enhance the final performance in particular object retrieval with compact codes.

Deep Metric Learning with Angular Loss

This paper proposes a novel angular loss, which takes angle relationship into account, for learning better similarity metric, and aims at constraining the angle at the negative point of triplet triangles.

Local Similarity-Aware Deep Feature Embedding

This paper introduces a Position-Dependent Deep Metric (PDDM) unit, which is capable of learning a similarity metric adaptive to local feature structure that can be used to select genuinely hard samples in a local neighborhood to guide the deep embedding learning in an online and robust manner.

Smart Mining for Deep Metric Learning

This paper proposes a novel deep metric learning method that combines the triplet model and the global structure of the embedding space and relies on a smart mining procedure that produces effective training samples for a low computational cost.

Improved Deep Metric Learning with Multi-class N-pair Loss Objective

This paper proposes a new metric learning objective called multi-class N-pair loss, which generalizes triplet loss by allowing joint comparison among more than one negative examples and reduces the computational burden of evaluating deep embedding vectors via an efficient batch construction strategy using only N pairs of examples.