Designing Category-Level Attributes for Discriminative Visual Recognition

By Felix X. Yu, Liangliang Cao, Rogério Schmidt Feris, John R. Smith, Shih-Fu Chang. 2013

220 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

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

Citations in the corpus

Transductive Multi-View Zero-Shot Learning (2015, background): “X S ∈ R n s × t and X T ∈ R n T × t denote the t − dimensional low-level feature vectors of auxiliary and target instances respectively. z S and z T are the auxiliary and target class label vectors.”
Evaluation of output embeddings for fine-grained image classification (2014, methodology): “Attributes have been widely used in the literature to aid classification and showed excellent results for image classification [16, 8, 13, 1, 34].”
Label-Embedding for Image Classification (2015, background): “Possible sources of prior information include attributes [30], [18], [43], [47], [46], semantic class taxonomies [46], [39], class-to-class similarities [47], [70], text features [43], [47], [46], [57], [20] or class co-occurrence statistics [37].”
An Empirical Study and Analysis of Generalized Zero-Shot Learning for Object Recognition in the Wild (2016, methodology): “Existing research on ZSL has been almost entirely focusing on this setting [4,10,11,12,13,14,15,8,16,17,18,19,20,21,22,23,24,25,26,27,28].”
Zero-Shot Object Detection: Joint Recognition and Localization of Novel Concepts (2020, background): “Based on how this relation is established, ZSL strategies can be categorized into four types. a) The ﬁrst type of methods attempt to predict class-speciﬁc semantic embeddings [37,54,24,63].”
Discovering Attribute Shades of Meaning with the Crowd (2015, methodology): “An approach that attempts to discover shades based on image clustering—or non-semantic attribute discovery such as (Parikh and Grauman 2011a; Mahajan et al. 2011; Duan et al. 2012; Rastegari et al. 2012; Sharmanska et al. 2012; Yu et al. 2013)—will be hard pressed to group images according to these ”
Zero-Shot Recognition via Structured Prediction (2016, background): “In the source domain, each class is represented by a single vector of side information such as attributes [3–7], language words/phrases [8–10], or even learned classifiers [11].”
Federated Learning with Only Positive Labels (2020, background): “In particular, similar to Proposition 1, Yu et al. (2013) shows that the classiﬁcation error can be bounded by the distance between data and positive label in the embedding space, and a measure of spreadout of the classes.”
Large-scale visual sentiment ontology and detectors using adjective noun pairs (2013, methodology): “Visual Feature Design: Following the feature design for state-of-the-art visual classification systems such as ObjectBank [38], we first include generic visual features: a 3× 256 dimension Color Histogram extracted from the RGB color channels, a 512 dimension GIST descriptor [40] that has been shown”
Predicting Visual Exemplars of Unseen Classes for Zero-Shot Learning (2016, methodology): “This is much more efficient than other ZSL algorithms that learn at the level of each individual training instance [11, 17, 28, 1, 47, 12, 39, 27, 13, 23, 2, 35, 49, 50, 21, 3].”
Improving Semantic Embedding Consistency by Metric Learning for Zero-Shot Classiffication (2016, methodology): “These issues have been addressed by automatically designing discriminative category-level attributes and using them for tasks of cross-category knowledge transfer, such as in the work of Yu et al. [18].”
Multi-modal Cycle-consistent Generalized Zero-Shot Learning (2018, background): “training and testing classes are typically acquired from other domains, such as visual features [4], text [5, 6, 3], or learned classifiers [7].”
Unsupervised Domain Adaptation for Zero-Shot Learning (2015, methodology): “Extensive experiments on four challenging object and action benchmark datasets [18, 34, 32] validate the advantages of the proposed model, and demonstrate that the proposed model outperforms the state-of-the-arts naive transfer learning based models [4, 18, 1, 10, 37, 15, 19, 20, 11, 2] when applied”
Zero-Shot Learning via Semantic Similarity Embedding (2015, background): “[42] proposed designing discriminative categorylevel attributes.”
Semi-supervised Vocabulary-Informed Learning (2016, unknown): “[46] A low-level 48.”
Learning Classifiers from Synthetic Data Using a Multichannel Autoencoder (2015, background): “Nevertheless, for certain tasks, such shared attributes [14], [12], [25], [10], [46] may simply be unavailable or nontrivial to deﬁne.”
Small Sample Learning in Big Data Era (2018, methodology): “, discriminative (Yu et al., 2013), interdependent (Jayaraman et al.”
Zero-Shot Learning via Joint Latent Similarity Embedding (2015, background): “In this paper we consider the source domain as a collection of class-level vectors, where each vector describes side information of one single class with, for instance, attributes [10, 19, 24, 27, 31], language words/phrases [4, 11, 34], or even learned classifiers [39].”
Zero-shot recognition with unreliable attributes (2014, methodology): “Whether the attribute vocabulary is hand-designed [8, 3, 19, 29, 23] or discovered [30, 11, 22], our approach assumes it is expressive enough to discriminate between the categories.”