GenZSL: Generative Zero-Shot Learning Via Inductive Variational Autoencoder

2025/05/17 by Shiming Chen, Dingjie Fu, Chen, Shiming +5 · 1 citation
Computer Science · #Computer Vision and Pattern Recognition (cs.CV) #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Face recognition and analysis #Generative Adversarial Networks and Image Synthesis #Machine Learning (cs.LG)

paper · pdf · doi:10.48550/arxiv.2505.11882

openalex publication_date 2025/05/17 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Remarkable progress in zero-shot learning (ZSL) has been achieved using generative models. However, existing generative ZSL methods merely generate (imagine) the visual features from scratch guided by the strong class semantic vectors annotated by experts, resulting in suboptimal generative performance and limited scene generalization. To address these and advance ZSL, we propose an inductive variational autoencoder for generative zero-shot learning, dubbed GenZSL. Mimicking human-level concept learning, GenZSL operates by inducting new class samples from similar seen classes using weak class semantic vectors derived from target class names (i.e., CLIP text embedding). To ensure the generation of informative samples for training an effective ZSL classifier, our GenZSL incorporates two key strategies. Firstly, it employs class diversity promotion to enhance the diversity of class semantic vectors. Secondly, it utilizes target class-guided information boosting criteria to optimize the model. Extensive experiments conducted on three popular benchmark datasets showcase the superiority and potential of our GenZSL with significant efficacy and efficiency over f-VAEGAN, e.g., 24.7% performance gains and more than 60× faster training speed on AWA2. Codes are available at https://github.com/shiming-chen/GenZSL.

Citations

• Visual-Augmented Dynamic Semantic Prototype for Generative Zero-Shot Learning
• Progressive Semantic-Guided Vision Transformer for Zero-Shot Learning
• Image-free Classifier Injection for Zero-Shot Classification
• Dataset Quantization
• ReCLIP: Refine Contrastive Language Image Pre-Training with Source Free Domain Adaptation
• Zero-Shot Learning by Harnessing Adversarial Samples
• Improving Zero-Shot Generalization for CLIP with Synthesized Prompts
• Self-regulating Prompts: Foundational Model Adaptation without Forgetting
• Evolving Semantic Prototype Improves Generative Zero-Shot Learning
• Synthetic Data from Diffusion Models Improves ImageNet Classification
• Your Diffusion Model is Secretly a Zero-Shot Classifier
• I2MVFormer: Large Language Model Generated Multi-View Document Supervision for Zero-Shot Image Classification
• Expanding Small-Scale Datasets with Guided Imagination
• Is synthetic data from generative models ready for image recognition?
• DUET: Cross-modal Semantic Grounding for Contrastive Zero-shot Learning
• Hierarchical Text-Conditional Image Generation with CLIP Latents
• MSDN: Mutually Semantic Distillation Network for Zero-Shot Learning
• High-Resolution Image Synthesis with Latent Diffusion Models
• High-Resolution Image Synthesis with Latent Diffusion Models
• TransZero: Attribute-guided Transformer for Zero-Shot Learning
• Fine-Grained Zero-Shot Learning with DNA as Side Information
• Learning to Prompt for Vision-Language Models
• FREE: Feature Refinement for Generalized Zero-Shot Learning
• Generative Models as a Data Source for Multiview Representation Learning
• Goal-Oriented Gaze Estimation for Zero-Shot Learning
• Counterfactual Zero-Shot and Open-Set Visual Recognition
• Learning Transferable Visual Models From Natural Language Supervision
• Attribute Prototype Network for Zero-Shot Learning
• Rethinking Generative Zero-Shot Learning: An Ensemble Learning Perspective for Recognising Visual Patches
• Invertible Zero-Shot Recognition Flows
• Leveraging the Invariant Side of Generative Zero-Shot Learning
• f-VAEGAN-D2: A Feature Generating Framework for Any-Shot Learning
• Semantic-Guided Multi-Attention Localization for Zero-Shot Learning
• Transductive Zero-Shot Learning with Visual Structure Constraint
• Generalized Zero- and Few-Shot Learning via Aligned Variational\n Autoencoders
• Generalized Zero-Shot Learning via Synthesized Examples
• Feature Generating Networks for Zero-Shot Learning
• mixup: Beyond Empirical Risk Minimization
• Zero-Shot Learning -- A Comprehensive Evaluation of the Good, the Bad and the Ugly
• Density Estimation Using Real NVP
• Latent Embeddings for Zero-shot Classification
• Human-level concept learning through probabilistic program induction

Cited by

• Interpretable Zero-Shot Learning with Locally-Aligned Vision-Language Model

Related

• Few-Shot Inspired Generative Zero-Shot Learning
• Continual Zero-Shot Learning through Semantically Guided Generative Random Walks
• End-to-end Generative Zero-shot Learning via Few-shot Learning
• Visual-Augmented Dynamic Semantic Prototype for Generative Zero-Shot Learning
• Boosting Generative Zero-Shot Learning by Synthesizing Diverse Features with Attribute Augmentation