Zero-Shot Learning -- The Good, the Bad and the Ugly

Abstract

Due to the importance of zero-shot learning, the number of proposed approaches has increased steadily recently. We argue that it is time to take a step back and to analyze the status quo of the area. The purpose of this paper is three-fold. First, given the fact that there is no agreed upon zero-shot learning benchmark, we first define a new benchmark by unifying both the evaluation protocols and data splits. This is an important contribution as published results are often not comparable and sometimes even flawed due to, e.g. pre-training on zero-shot test classes. Second, we compare and analyze a significant number of the state-of-the-art methods in depth, both in the classic zero-shot setting but also in the more realistic generalized zero-shot setting. Finally, we discuss limitations of the current status of the area which can be taken as a basis for advancing it.

Accepted as a full paper in IEEE Computer Vision and Pattern Recognition (CVPR) 2017. We introduce Proposed Split Version 2.0 (Please download it from the project page)

References in corpus (1)

• Zero-Shot Learning Through Cross-Modal Transfer

Cited by in corpus (49)

• Deep Learning for Anomaly Detection: A Survey
• Leveraging the Invariant Side of Generative Zero-Shot Learning
• Preserving Semantic Relations for Zero-Shot Learning
• A causal view of compositional zero-shot recognition
• On the Effectiveness of Transfer Learning for Code Search
• Zero-shot Recognition via Semantic Embeddings and Knowledge Graphs
• Probabilistic AND-OR Attribute Grouping for Zero-Shot Learning
• Partially-Supervised Image Captioning
• Towards Discriminability and Diversity: Batch Nuclear-norm Maximization under Label Insufficient Situations
• DSGAN: Generative Adversarial Training for Distant Supervision Relation Extraction
• Multi-Head Self-Attention via Vision Transformer for Zero-Shot Learning
• Attributes' Importance for Zero-Shot Pose-Classification Based on Wearable Sensors
• Zero-Shot Visual Recognition using Semantics-Preserving Adversarial Embedding Networks
• Learning Class Prototypes via Structure Alignment for Zero-Shot Recognition
• Episode-based Prototype Generating Network for Zero-Shot Learning
• Learning the Redundancy-free Features for Generalized Zero-Shot Object Recognition
• Visual Data Synthesis via GAN for Zero-Shot Video Classification
• Open-World Visual Recognition Using Knowledge Graphs
• Weak-shot Fine-grained Classification via Similarity Transfer
• KG-GAN: Knowledge-Guided Generative Adversarial Networks
• Dizygotic Conditional Variational AutoEncoder for Multi-Modal and Partial Modality Absent Few-Shot Learning
• A Large-scale Attribute Dataset for Zero-shot Learning
• Classifier and Exemplar Synthesis for Zero-Shot Learning
• Adaptive Cross-Modal Few-Shot Learning
• Composed Variational Natural Language Generation for Few-shot Intents
• Zero and Few Shot Learning with Semantic Feature Synthesis and Competitive Learning
• SSSE: Efficiently Erasing Samples from Trained Machine Learning Models
• Zero-Shot Learning via Category-Specific Visual-Semantic Mapping
• Cycle-consistent Conditional Adversarial Transfer Networks
• Bi-Adversarial Auto-Encoder for Zero-Shot Learning
• Generalized Zero-Shot Learning via VAE-Conditioned Generative Flow
• Generalized Zero-Shot Learning for Action Recognition with Web-Scale Video Data
• Alternative Semantic Representations for Zero-Shot Human Action Recognition
• Global Semantic Consistency for Zero-Shot Learning
• Alleviating Feature Confusion for Generative Zero-shot Learning
• Learning Local Shape Descriptors from Part Correspondences With Multi-view Convolutional Networks
• Spatial Priming for Detecting Human-Object Interactions
• Unsupervised Open Domain Recognition by Semantic Discrepancy Minimization
• A Biologically Inspired Feature Enhancement Framework for Zero-Shot Learning
• Fine-grained Classification using Heterogeneous Web Data and Auxiliary Categories
• Class Knowledge Overlay to Visual Feature Learning for Zero-Shot Image Classification
• Few-Shot Learning with Per-Sample Rich Supervision
• Beyond Attributes: Adversarial Erasing Embedding Network for Zero-shot Learning
• Few-Shot Adaptation for Multimedia Semantic Indexing
• Zero-Shot Learning Based on Knowledge Sharing
• On zero-shot recognition of generic objects
• Learning Graph-Based Priors for Generalized Zero-Shot Learning
• Label Embedding with Partial Heterogeneous Contexts
• Recognizing Part Attributes with Insufficient Data