Learning to share visual appearance for multiclass object detection

By Ruslan Salakhutdinov, Antonio Torralba, Josh Tenenbaum. 2011

362 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

Scene Graph Generation by Iterative Message Passing (2017, background): “Visual scene understanding often harnesses the statistical patterns of object co-occurrence [11, 22, 30, 35] as well as spatial layout [2, 9].”
Large-Scale Long-Tailed Recognition in an Open World (2019, background): “Arising from long-tail distributions of natural data, it has been extensively studied [41, 61, 3, 30, 62, 34, 29, 49, 6].”
Visual Relationship Detection with Language Priors (2016, background): “There has been a series of work related to improving object detection by leveraging object co-occurrence statistics [9,10,11,12,13,14].”
The Devil is in the Tails: Fine-grained Classification in the Wild (2017, background): “In a nutshell: the world is long-tailed, as previously noted in the context of subcategories and object views [43, 64].”
An Empirical Study and Analysis of Generalized Zero-Shot Learning for Object Recognition in the Wild (2016, background): “It is well-known, however, that object frequencies in natural images follow longtailed distributions [1,2,3].”
Recent Advances in Open Set Recognition: A Survey (2018, background): “On the other hand, the object frequencies in natural world follow long-tailed distributions [28], [29], meaning that KKCs are more common than UKCs.”
Multi-Task CNN Model for Attribute Prediction (2015, background): “and unlike [44], the mixed norm regularizations require re-representing the 2-norm into its dual form as discussed in [21].”
Learning Semantic Image Representations at a Large Scale (2014, background): “[96] proposed learning a set of object classifiers with regularization using hierarchical knowledge, which improves the classification of objects at the leaves of the hierarchy.”
Learning Classifiers from Synthetic Data Using a Multichannel Autoencoder (2015, background): “In terms of quantity, it has been shown that the amount of available training data, per object class, roughly follows a Zipf distribution [35].”
Discriminative Transfer Learning with Tree-based Priors (2013, background): “Most similar to our work is [18] which defined a generative prior over the classifier parameters and a prior over the tree structures to identify relevant categories.”
From categories to subcategories: Large-scale image classification with partial class label refinement (2015, background): “Class hierarchy can also be implictly learned to share model parameters between parent and children nodes [29] or to speed up the classification by stacking classifiers [22].”
Solving Long-tailed Recognition with Deep Realistic Taxonomic Classifier (2020, methodology): “s for each input example. This is addressed with a novel dynamic predictor synthesis procedure inspired by parameter inheritance, a regularization strategy commonly used in hierarchical classication [51,52]. To the best of our knowledge, these contributions enable the rst implementation of RTC with ”
Write a Classifier: Zero-Shot Learning Using Purely Textual Descriptions (2013, background): “This is reflected in the number of images per category available for training in most object categorization datasets, which, as pointed out in [27], shows a Zipf distribution.”
Which Looks Like Which: Exploring Inter-class Relationships in Fine-Grained Visual Categorization (2014, positive): “For example, adopting class taxonomies or hierarchies [6,23] may help train better prediction models by sharing appearance [20,30], visual parts [33], or classifiers [4].”
What leads to generalization of object proposals? (2020, background): “Generalizing localization across multiple classes: The idea of different object classes sharing the same structure has been exploited in building detection models for a long time[5,22,23,29,34].”
ImageNet Auto-Annotation with Segmentation Propagation (2014, methodology): “Many methods use the parameters of the source classiﬁers as priors for the target model [5,6,25,55, 63].”
Towards more human-like concept learning in machines : compositionality, causality, and learning-to-learn (2014, background): “For discriminative models such as convolutional networks and classic object recognition approaches, learning-to-learn can occur through the sharing of features between the classifiers learned for old objects and the classifiers learned for new objects (Torralba, Murphy, & Freeman, 2007; Salakhutdino”