Preserving Semantic Relations for Zero-Shot Learning

Abstract

Zero-shot learning has gained popularity due to its potential to scale recognition models without requiring additional training data. This is usually achieved by associating categories with their semantic information like attributes. However, we believe that the potential offered by this paradigm is not yet fully exploited. In this work, we propose to utilize the structure of the space spanned by the attributes using a set of relations. We devise objective functions to preserve these relations in the embedding space, thereby inducing semanticity to the embedding space. Through extensive experimental evaluation on five benchmark datasets, we demonstrate that inducing semanticity to the embedding space is beneficial for zero-shot learning. The proposed approach outperforms the state-of-the-art on the standard zero-shot setting as well as the more realistic generalized zero-shot setting. We also demonstrate how the proposed approach can be useful for making approximate semantic inferences about an image belonging to a category for which attribute information is not available.

arXiv commentCVPR 2018

References in corpus (13)

• Adam: A Method for Stochastic Optimization
• Efficient Estimation of Word Representations in Vector Space
• Deep Residual Learning for Image Recognition
• Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)
• Zero-Shot Learning by Convex Combination of Semantic Embeddings
• Zero-Shot Learning -- A Comprehensive Evaluation of the Good, the Bad and the Ugly
• Predicting Deep Zero-Shot Convolutional Neural Networks using Textual Descriptions
• Semantic Autoencoder for Zero-Shot Learning
• Zero-Shot Learning via Semantic Similarity Embedding
• Zero-Shot Learning -- The Good, the Bad and the Ugly
• Ridge Regression, Hubness, and Zero-Shot Learning
• Semantically Consistent Regularization for Zero-Shot Recognition
• Zero-Shot Learning via Joint Latent Similarity Embedding

Cited by in corpus (28)

• SG-One: Similarity Guidance Network for One-Shot Semantic Segmentation
• Transductive Zero-Shot Learning with Visual Structure Constraint
• Multi-modal Cycle-consistent Generalized Zero-Shot Learning
• Dual Progressive Prototype Network for Generalized Zero-Shot Learning
• Contrastive Embedding for Generalized Zero-Shot Learning
• SFE-GACN: A Novel Unknown Attack Detection Method Using Intra Categories Generation in Embedding Space
• Zero-Shot Learning with Common Sense Knowledge Graphs
• Generalized Zero-Shot Recognition based on Visually Semantic Embedding
• Adaptive Confidence Smoothing for Generalized Zero-Shot Learning
• A Boundary Based Out-of-Distribution Classifier for Generalized Zero-Shot Learning
• Domain segmentation and adjustment for generalized zero-shot learning
• Active learning for interactive satellite image change detection
• Towards Effective Deep Embedding for Zero-Shot Learning
• Generative Dual Adversarial Network for Generalized Zero-shot Learning
• Bi-Adversarial Auto-Encoder for Zero-Shot Learning
• Global Semantic Consistency for Zero-Shot Learning
• ATZSL: Defensive Zero-Shot Recognition in the Presence of Adversaries
• Domain-Specific Embedding Network for Zero-Shot Recognition
• Hierarchical Prototype Learning for Zero-Shot Recognition
• Bi-Semantic Reconstructing Generative Network for Zero-shot Learning
• Learning Classifiers for Domain Adaptation, Zero and Few-Shot Recognition Based on Learning Latent Semantic Parts
• Attribute-Induced Bias Eliminating for Transductive Zero-Shot Learning
• Zero-Shot Recognition through Image-Guided Semantic Classification
• Convolutional Prototype Learning for Zero-Shot Recognition
• Using Fictitious Class Representations to Boost Discriminative Zero-Shot Learners
• Visual and Semantic Prototypes-Jointly Guided CNN for Generalized Zero-shot Learning
• Toward Zero-Shot Unsupervised Image-to-Image Translation
• Beyond Attributes: Adversarial Erasing Embedding Network for Zero-shot Learning