Multi-attribute spaces: Calibration for attribute fusion and similarity search

By Walter J. Scheirer, Neeraj Kumar, Peter N. Belhumeur, Terrance E. Boult. 2012

197 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Media Technology, Information Retrieval & Search

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In Media Technology

In Information Retrieval & Search

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, background): “caption generation [28, 39], face recognition [48, 6], image retrieval [49, 11], action recognition [33, 57] and image classification [29, 2].”
Latent Embeddings for Zero-Shot Classification (2016, background): “They are typically collected through costly human annotation [7, 17, 28] and have shown promising results [1, 3, 6, 20, 22, 33, 34, 40] in various computer vision problems.”
Label-Embedding for Image Classification (2015, background): “We now review related work on attributes, zero-shot learning and label embedding, three research areas which strongly overlap.”
Multi-Task CNN Model for Attribute Prediction (2015, background): “However, most of the existing literature employs discriminative classifiers independently to predict the semantic attributes of low-level image features [13], [30], [46].”
Learning Attributes Equals Multi-Source Domain Generalization (2016, background): “Moreover, attributes offer a natural humancomputer interaction channel for visual recognition with humans in the loop [7], relevance feedback in image retrieval [42, 64, 55, 60, 81, 40, 58, 25, 72], and active learning [41, 56, 5, 46, 43].”
The Extreme Value Machine (2015, background): “With the observation that the tails of any distance or similarity score distribution must always follow an EVT distribution [23], highly accurate probabilistic calibration models became possible, leading to strong empirical results for multi-biometric fusion [24], describable visual attributes [25],”
Attribute relation learning for zero-shot classification (2014, background): “search and retrieval [15–18] has been developed with promising results.”
Transductive Multi-view Embedding for Zero-Shot Recognition and Annotation (2014, background): “Most existing studies [18, 17, 25, 26, 30, 37, 1] assume that an exhaustive ontology of attributes has been manually specified at either the class or instance level.”
Data-Driven 3D Primitives for Single Image Understanding (2013, unknown): “[25], which requires only negative data.”
Best of both worlds: Human-machine collaboration for object annotation (2015, background): “The true answer is then determined to be “no” to three questions: (1) when asked if C is contained in the image (verify-image), (2) when asked if C is contained in any box B (verify-box), and (3) when asked if there are more instances of class C in the image (verify-cover).”
Multi-class Open Set Recognition Using Probability of Inclusion (2014, methodology): “For multi-class recognition problems in computer vision, posterior probabilities are widely used to make decisions in applications such as pedestrian classification and orientation estimation [18], image retrieval [15], attribute fusion [42], part-based human tracking [47], large-scale multi-class o”
Learning and Calibrating Per-Location Classifiers for Visual Place Recognition (2013, methodology): “(Scheirer et al. 2012) who develop a classifier calibration method for face attribute similarity search.”
Attribute-Based Classification for Zero-Shot Visual Object Categorization (methodology): “Scheirer et al. [44] introduced an alternative technique for turning the output of attribute classifiers into probability estimates based on extremal value theory.”
Deep Classifiers from Image Tags in the Wild (2015, methodology): “We tested the calibration method from [27], but found it degraded performance.”
Towards Transparent Systems: Semantic Characterization of Failure Modes (2014, background): “Predicting failure: Methods that predict overall performance of a system on a collection of test images by analyzing statistics of the test data or postrecognition scores [10–15] are not applicable to our goal of identifying specific failure modes of the system, and semantically characterizing them.”
Predicting Failures of Vision Systems (2014, background): “The analysis can be used to automatically determine thresholds to make match / nonmatch decisions [50], intelligent fusion schemes [47] [48], etc. ALERT differs from this line of work in two prominent ways (1) ALERT does not require the output of BASESYS on a test instance to predict its…”
Demographic Estimation from Face Images: Human vs. Machine Performance (2015, background): “There has been a growing interest in automatic extraction of demographic information from face images or videos, due to many emerging applications [2], [4], [5], [6].”