Zero-Shot Learning — The Good, the Bad and the Ugly
By Yongqin Xian, Bernt Schiele, Zeynep Akata. 2017
969 citations recorded by the source corpus.
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.
Subfields: General Dentistry, Surgery, Meta & Continual Learning
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In General Dentistry
- Fei Sha
- Ashish Mishra
- Hexiang Hu
- Soma Biswas
- Jaya Kalla
- Mert Bulent Sariyildiz
- De-chuan Zhan
- R. G. Cinbis
In Surgery
- Fei Sha
- Ashish Mishra
- Hexiang Hu
- Soma Biswas
- Jaya Kalla
- Mert Bulent Sariyildiz
- De-chuan Zhan
- R. G. Cinbis
In Meta & Continual Learning
- Han-Jia Ye
- De-chuan Zhan
- K. J. Joseph
- Vineeth Nallure Balasubramanian
- Salman Hameed Khan
- Wei-Lun Chao
- Venkatesh Saligrama
- Shafin Rahman
Citations in the corpus
- Unified Contrastive Learning in Image-Text-Label Space (2022, methodology): “In this sense, our work is also related to conventional zero-shot classification [9, 28, 45, 62, 65, 66].”
- S3C: Self-Supervised Stochastic Classifiers for Few-Shot Class-Incremental Learning (2023, methodology): “To better capture this performance balance, inspired from [39], recent FSCIL works [4,9] propose to use the Harmonic Mean (HM) of the performance of the previous and new classes at the end of each incremental task.”
- Novel Class Discovery without Forgetting (2022, background): “Zero-shot learning methods [54,42] require prior knowledge of extra semantic attribute information about the unlabeled classes.”
- Learning to Reconstruct Shapes from Unseen Classes (2018, methodology): “We refer readers to the review article [Xian et al., 2017] for a comprehensive list.”
- Learning Adaptive Classifiers Synthesis for Generalized Few-Shot Learning (2019, methodology): “Zero-shot learning (ZSL) [2, 8, 29, 68] is a popular idea for addressing learning without labeled data.”
- Conditional Prompt Learning for Vision-Language Models (2022, background): “, to recognize novel classes by training only on base classes [3,51,54,57].”
- XOC: Explainable Observer-Classifier for Explainable Binary Decisions (2019, methodology): “As the zero-shot learning model, we use Structured Joint Embedding (SJE) (Akata et al., 2015), the best zero-shot-learning performance on AWA2 according to Xian et al. (2017).”
- Don’t Even Look Once: Synthesizing Features for Zero-Shot Detection (2019, background): “As such, ZSL exhibits bias towards unseen classes, and GZSL evaluation attempts to rectify it by evaluating on both seen and unseen objects at test-time [3, 43, 12, 17, 15, 46, 37, 47].”
- Domain-Invariant Projection Learning for Zero-Shot Recognition (2018, methodology): “Taking the same ‘pure’ ZSL setting as in [60, 29], we remove the overlapped ImageNet ILSVRC2012 1K classes from the test set of unseen classes for the four medium-scale datasets.”
- Probabilistic Embeddings for Cross-Modal Retrieval (2021, neutral): “[57], where 150 classes are used for training and validation, and the remaining 50 classes are used for the test.”
- Doubly Right Object Recognition: A Why Prompt for Visual Rationales (2022, methodology): “[58] used the visual attributes for animal classification and [23] used them for face verification.”