Metric Embedding Autoencoders for Unsupervised Cross-Dataset Transfer Learning

TL;DR

This paper introduces a deep autoencoder-based transfer learning model for person re-identification that improves cross-dataset performance without relying on co-training with non-deep models.

Contribution

It presents a novel deep autoencoder architecture with metric embedding for unsupervised cross-dataset transfer learning in person Re-ID.

Findings

Outperforms baseline models in transfer learning tasks

Effective unsupervised training without co-training

Improves cross-dataset generalization in person Re-ID

Abstract

Cross-dataset transfer learning is an important problem in person re-identification (Re-ID). Unfortunately, not too many deep transfer Re-ID models exist for realistic settings of practical Re-ID systems. We propose a purely deep transfer Re-ID model consisting of a deep convolutional neural network and an autoencoder. The latent code is divided into metric embedding and nuisance variables. We then utilize an unsupervised training method that does not rely on co-training with non-deep models. Our experiments show improvements over both the baseline and competitors' transfer learning models.