Rethinking Feature Discrimination and Polymerization for Large-scale Recognition

TL;DR

This paper introduces the COCO algorithm, which enhances feature discrimination and polymerization in deep networks for large-scale recognition by optimizing cosine similarity, leading to improved classification performance.

Contribution

The paper proposes the COCO algorithm that jointly optimizes cosine similarity and class centroids for better feature discrimination and polymerization in large-scale recognition tasks.

Findings

Effective on multiple benchmarks for small and large-scale recognition

Improves intra-class feature compactness and inter-class separability

Achieves stable convergence during end-to-end training

Abstract

Feature matters. How to train a deep network to acquire discriminative features across categories and polymerized features within classes has always been at the core of many computer vision tasks, specially for large-scale recognition systems where test identities are unseen during training and the number of classes could be at million scale. In this paper, we address this problem based on the simple intuition that the cosine distance of features in high-dimensional space should be close enough within one class and far away across categories. To this end, we proposed the congenerous cosine (COCO) algorithm to simultaneously optimize the cosine similarity among data. It inherits the softmax property to make inter-class features discriminative as well as shares the idea of class centroid in metric learning. Unlike previous work where the center is a temporal, statistical variable within one mini-batch during training, the formulated centroid is responsible for clustering inner-class features to enforce them polymerized around the network truncus. COCO is bundled with discriminative training and learned end-to-end with stable convergence. Experiments on five benchmarks have been extensively conducted to verify the effectiveness of our approach on both small-scale classification task and large-scale human recognition problem.