Safety in numbers: Learning categories from few examples with multi model knowledge transfer

By Tatiana Tommasi, Francesco Orabona, Barbara Caputo. 2010

231 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

A Comprehensive Survey on Transfer Learning (2019, methodology): “In [130], an approach that is referred to as multimodel knowledge transfer (MMKL) is proposed.”
Domain Adaptation for Visual Applications: A Comprehensive Survey (2017, methodology): “e seen as a sort of two-task learning problem where, in particular, learning the source speciﬁc task helps learning the target speciﬁc task. Furthermore, in the case of multi-source domain adaptation [230, 231, 89, 87, 232, 86, 28, 233, 62, 77] different source speciﬁc tasks are jointly exploited in”
Learning and Transferring Mid-level Image Representations Using Convolutional Neural Networks (2014, methodology): “Much of this progress has been enabled by the development of robust image descriptors such as SIFT [32] and HOG [8], bagof-features image representations [7, 26, 36, 45] as well as deformable part models [14].”
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].”
Asymmetric multi-task learning based on task relatedness and loss (2016, methodology): “Transfer learning Our regularization term closely resembles the least square-based regularization term in Tommasi et al. (2010), which is used to perform multi-model transfer for a least-square one-vs-all SVM.”
From N to N+1: Multiclass Transfer Incremental Learning (2013, methodology): “Success in this setting is deﬁned as optimizing the accuracy of the dog detector, with a minimal number of annotated training samples (Figure 1, left).”
Cross-Modal Supervision for Learning Active Speaker Detection in Video (2016, background): “Transfer Learning [18,19,20], a related problem, is about using the information available from the source data to aid the learning of the target classiﬁer utilizing only a small number of target training samples.”
Learning to share visual appearance for multiclass object detection (2011, background): “Other models learn the relations between categories [30, 21, 22, 1, 25, 13, 29].”
Incremental robot learning of new objects with fixed update time (2016, background): “Future research will focus on strategies to exploit knowledge of known classes to improve classification accuracy on new ones, following recent work [20], [21], [22], [23].”
Shifting Weights: Adapting Object Detectors from Image to Video (2012, methodology): “More similar to our method are approaches based on optimizing Support Vector Machine (SVM) related objectives [19] [20] [21] [22] [23] [24] or joint cost functions [25], that treat the features as fixed and seek to adapt parameters of the classifier from source to target domain.”
Error-Driven Incremental Learning in Deep Convolutional Neural Network for Large-Scale Image Classification (2014, methodology): “The second is when there are existing classiﬁers that are related to the new classes to be learned [1, 7, 8, 9, 15, 16, 21, 22].”
Label Efficient Learning of Transferable Representations acrosss Domains and Tasks (2017, background): “In computer vision, examples of transfer learning include [1, 31, 61] which try to overcome the deficit of training samples for some categories by adapting classifiers trained for other categories [43].”
Incremental Learning of NCM Forests for Large-Scale Image Classification (2014, methodology): “It can be implemented via additional SVM constraints [22], regularizers [2], transferred features [21], shared deformable parts [16] or iteratively learned object localization based on global annotations [12].”
Weakly Supervised Object Localization Using Things and Stuff Transfer (2017, unknown): “It is intensively studied in image classification, segmentation and object detection [32, 33, 34, 35, 36, 37, 38].”
The Multiverse Loss for Robust Transfer Learning (2015, background): “. Introduction One of the hallmarks of the recent success of deep learning methods in computer vision is the ability to learn effective representations in one domain and apply these on another domain [15,7,6,39,22,34,16]. The source and target domains might differ in the underlying probability distr”