Latent Embedding Feedback and Discriminative Features for Zero-Shot Classification

Document Type

Conference Proceeding

Source of Publication

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

Publication Date

1-1-2020

Abstract

© 2020, Springer Nature Switzerland AG. Zero-shot learning strives to classify unseen categories for which no data is available during training. In the generalized variant, the test samples can further belong to seen or unseen categories. The state-of-the-art relies on Generative Adversarial Networks that synthesize unseen class features by leveraging class-specific semantic embeddings. During training, they generate semantically consistent features, but discard this constraint during feature synthesis and classification. We propose to enforce semantic consistency at all stages of (generalized) zero-shot learning: training, feature synthesis and classification. We first introduce a feedback loop, from a semantic embedding decoder, that iteratively refines the generated features during both the training and feature synthesis stages. The synthesized features together with their corresponding latent embeddings from the decoder are then transformed into discriminative features and utilized during classification to reduce ambiguities among categories. Experiments on (generalized) zero-shot object and action classification reveal the benefit of semantic consistency and iterative feedback, outperforming existing methods on six zero-shot learning benchmarks. Source code at https://github.com/akshitac8/tfvaegan.

ISBN

9783030585419

ISSN

0302-9743

Publisher

Springer International Publishing

Volume

12367 LNCS

First Page

479

Last Page

495

Disciplines

Computer Sciences

Keywords

Feature synthesis, Generalized zero-shot classification

Indexed in Scopus

yes

Open Access

yes

Open Access Type

Green: A manuscript of this publication is openly available in a repository

Link to Full Text

Share

COinS