• Home
  • Search
  • GSMFlow: Generation Shifts Mitigating Flow for Generalized Zero-Shot Learning
  • Open Access IconOpen Access
  • Cite Icon15
  • https://doi.org/10.1109/tmm.2022.3190678Copy DOI Icon

GSMFlow: Generation Shifts Mitigating Flow for Generalized Zero-Shot Learning

Show More
  • Abstract
  • Literature Map
  • References
  • Citations
  • Similar Papers
Abstract

Generalized Zero-Shot Learning (GZSL) aims to recognize images not only for seen classes but also for unseen ones by transferring semantic-visual relationships from the seen to the unseen classes. It is an intuitive solution to take the advantage of generative models to hallucinate realistic unseen samples based on the knowledge learned from the seen classes. However, due to the generation shifts, the synthesized samples by most existing methods may drift from the real distribution of the unseen data. To address this issue, we propose a novel flow-based generative framework that consists of multiple conditional affine coupling layers for learning unseen data generation. Specifically, we investigate and address three essential problems that trigger the generation shifts, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e.</i> , <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">semantic inconsistency</i> , <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">variance collapse</i> , and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">structure disorder</i> . First, to improve the reflection of the semantic information in the generated samples, we proactively embed the semantic information into the transformation in each conditional affine coupling layer. Second, to promote the intrinsic feature variance of the unseen classes, we introduce a boundary sample mining strategy with entropy maximization to discover ambiguous visual variants of semantic prototypes and hereby calibrate the decision boundary of the classifiers. Third, a relative positioning strategy is proposed to revise the attribute embeddings, guiding which to fully preserve the inter-class geometric structure and further avoid structure disorder in the semantic space. Extensive experimental results on four GZSL benchmark datasets demonstrate that GSMFlow achieves the state-of-the-art performance on GZSL.

Similar Papers
  • Research Article
  • Citations18

GAN-MVAE: A discriminative latent feature generation framework for generalized zero-shot learning

  • Mar 01, 2022
  • Pattern Recognition Letters
  • Peirong Ma +3
  • Research Article
  • Citations7

Open zero-shot learning via asymmetric VAE with dissimilarity space

  • Jul 20, 2023
  • Information Sciences
  • Zhibo Zhai +2
  • Conference Article
  • Citations16

Generalized Zero-Shot Learning using Generated Proxy Unseen Samples and Entropy Separation

  • Oct 12, 2020
  • Omkar Gune +3
  • Research Article
  • Citations13

Domain-Aware Prototype Network for Generalized Zero-Shot Learning

  • May 01, 2024
  • IEEE Transactions on Circuits and Systems for Video Technology
  • Yongli Hu +4
  • Conference Article
  • Citations16

VS-Boost: Boosting Visual-Semantic Association for Generalized Zero-Shot Learning

  • Aug 01, 2023
  • Xiaofan Li +6
  • Research Article
  • Citations35

Generative Mixup Networks for Zero-Shot Learning.

  • Mar 01, 2025
  • IEEE transactions on neural networks and learning systems
  • Bingrong Xu +3
  • Report Series

From Classical to Generalized Zero-Shot Learning: a Simple Adaptation Process

  • Oct 29, 2018
  • EasyChair preprint
  • Yannick Le Cacheux +2
  • Research Article
  • Citations12

A Semantic Encoding Out-of-Distribution Classifier for Generalized Zero-Shot Learning

  • Jan 01, 2021
  • IEEE Signal Processing Letters
  • Jiayu Ding +2
  • PDF
  • Research Article
  • Citations5

Co-GZSL: Feature Contrastive Optimization for Generalized Zero-Shot Learning

  • Mar 12, 2024
  • Neural Processing Letters
  • Qun Li +3
  • Conference Article
  • Citations118

Semantics Disentangling for Generalized Zero-Shot Learning

  • Oct 01, 2021
  • Zhi Chen +6
  • PDF
  • Research Article
  • Citations1

GT-GAN: A General Transductive Zero-Shot Learning Method Based on GAN

  • Jan 01, 2020
  • IEEE Access
  • Junhao Dong +3
  • PDF
  • Research Article
  • Citations13

Cooperative Coupled Generative Networks for Generalized Zero-Shot Learning

  • Jan 01, 2020
  • IEEE Access
  • Liang Sun +3
  • PDF
  • Research Article
  • Citations8

Discriminative Embedding Autoencoder With a Regressor Feedback for Zero-Shot Learning

  • Jan 01, 2020
  • IEEE Access
  • Ying Shi +1
  • Research Article
  • Citations10

Dual Projective Zero-Shot Learning Using Text Descriptions

  • Jan 05, 2023
  • ACM Transactions on Multimedia Computing, Communications, and Applications
  • Yunbo Rao +4
  • Conference Article
  • Citations2

Dual Prototype Relaxation for Generalized Zero Shot Learning

  • Jul 05, 2021
  • Jie Zhang +2
Cactus Communications logo

Copyright 2026 Cactus Communications. All rights reserved.