Transductive Multi-view Embedding for Zero-Shot Recognition and Annotation

By Yanwei Fu, Timothy M. Hospedales, Tao A. Xiang, Zhenyong Fu, Shaogang Gong. 2014

250 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Representation & Self-Supervised Learning

Citations in the corpus

Generative Adversarial Text to Image Synthesis (2016, background): “…the object category encoded into a vector (Farhadi et al., 2009; Kumar et al., 2009; Parikh & Grauman, 2011; Lampert et al., 2014), in particular to enable zero-shot visual recognition (Fu et al., 2014; Akata et al., 2015), and recently for conditional image generation (Yan et al., 2015).”
Learning Deep Representations of Fine-Grained Visual Descriptions (2016, methodology): “For instance on the Caltech-UCSD birds database (CUB) [46], previous zero-shot learning approaches [13, 2, 3] have used human-encoded attributes [24], or simplified language models such as bag-of-words [16], WordNet-hierarchyderived features [29], and neural word embeddings such as Word2Vec [28] and”
Learning to Compare: Relation Network for Few-Shot Learning (2017, background): “cent methods [48, 12, 13] are tranductive in that they use all test data at once for model training, which gives them a big advantage at the cost of making a very strong assumption that may not be met in practical applications, so we do not compare with them here.”
Transductive Multi-View Zero-Shot Learning (2015, methodology): “The same Bayesian model averaging [14] can be used here to estimate these prior probabilities.”
Evaluation of output embeddings for fine-grained image classification (2014, background): “Learning to classify in the absence of labeled data (zeroshot learning) [58, 45, 25, 29, 2, 37, 34, 17] is a challenging problem, and achieving better-than-chance performance requires structure in the output space.”
Domain Adaptation for Visual Applications: A Comprehensive Survey (2017, methodology): “ask is deduced from previous tasks without using any training data for the current task. To address zero-shot learning, the methods rely either on nameable image characteristics and semantic concepts [238, 239, 240, 241], or on latent topics discovered by the system directly from the data [242, 243,”
Generalized Zero- and Few-Shot Learning via Aligned Variational Autoencoders (2018, methodology): “In doing so, our method is less biased towards learning the distribution of the seen class image features, which is known as the projection domain shift problem [7].”
Learning a Deep Embedding Model for Zero-Shot Learning (2016, methodology): “Most state-of-the-arts ZSL models [11, 13, 2, 3, 37, 47, 22] use deep CNN features for visual feature representation; the features are extracted with pretrained CNN models.”
Synthesized Classifiers for Zero-Shot Learning (2016, methodology): “More recent studies show that nearest neighbor classifiers in the semantic space are very effective [10, 11, 12, 13, 30].”
Semantic Autoencoder for Zero-Shot Learning (2017, background): “In the semantic embedding space, the names of both seen and unseen classes are embedded as vectors called class prototypes [20].”
Classifier and Exemplar Synthesis for Zero-Shot Learning (2018, unknown): “Some propose to use canonical correlation analysis (CCA) [18,19,42].”
Spatial-Aware Object Embeddings for Zero-Shot Localization and Classification of Actions (2017, unknown): “using transductive learning [9, 58] and domain adaption [23, 57].”
Interestingness Prediction by Robust Learning to Rank (2014, background): “In a broader sense of attributes [26,11,12,27,13] interestingness can be considered as one type of relative attributes [35], although those attributes, such as how smiling a person is, are much less subjective.”
A Unified Perspective on Multi-Domain and Multi-Task Learning (2014, methodology): “Going beyond information sharing for known tasks, an exciting related paradigm for task-transfer is “zero-shot” learning (ZSL) (Larochelle et al., 2008; Lampert et al., 2009; Fu et al., 2014).”
Gradient Matching Generative Networks for Zero-Shot Learning (2019, background): “Alternatively, transductive approaches have been proposed to benefit from unlabeled data [43, 23, 44, 39].”
Zero-Shot Learning via Joint Latent Similarity Embedding (2015, methodology): “(iv) Other methods: Less related to our method includes approaches based on semantic transfer propagation [30], transductive multi-view embedding [12], random forest approach [15], and semantic manifold distance [13].”
Expanding object detector's Horizon: Incremental learning framework for object detection in videos (2015, background): “The choice of proposing an embedding-based detection paradigm, over the more traditional SVMs or latent SVM, stems from flexibility and scalability of such models, their ability to generalize with little to no data [24], as well as their state-of-the-art performance on large-scale categorization tas”
Unsupervised Domain Adaptation for Zero-Shot Learning (2015, methodology): “The solution offered in [10] is a heuristic one-step self-training strategy to pull the prototype towards its closet data points (not necessary from the same class) followed by a multi-view embedding based on canonical correlation analysis (CCA) to align different semantic spaces with the low-level ”
Transductive Learning for Zero-Shot Object Detection (2019, background): “Given the challenging nature of the ZSD problem, it is of great interest to study how transductive settings can help in dealing with domain-shift [4] and model-bias [2] problems.”