Towards good practice in large-scale learning for image classification

By Zeynep Akata, Florent Perronnin, Zaïd Harchaoui, Cordelia Schmid. 2012

206 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Optimization & Training

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 Optimization & Training

Citations in the corpus

ImageNet Large Scale Visual Recognition Challenge (2014, background): “…Sande, Amir Habibian, Cees G. M. Snoek (Sanchez and Perronnin, 2011; Scheirer et al., 2012) XRCE 27.1 - Xerox Research Centre Europe†, LEAR INRIA ‡ Florent Perronnin†, Zeynep Akata†‡, Zaid Harchaoui‡, Cordelia Schmid‡ (Perronnin et al., 2012) Table 5 Teams participating in ILSVRC2010-2012,…”
ReferItGame: Referring to Objects in Photographs of Natural Scenes (2014, background): “Recent advances in automatic computer vision methods have started to make technologies for recognizing thousands of object categories a near reality (Perronnin et al., 2012; Deng et al., 2012; Deng et al., 2010; Krizhevsky et al., 2012).”
Billion-scale semi-supervised learning for image classification (2019, methodology): “Inspired by a prior work [32] on large-scale supervised learning, we summarize these recommendations in Table 1.”
Return of the Devil in the Details: Delving Deep into Convolutional Nets (2014, methodology): “We also determine scenario-speciﬁc best-practices , improving the ones in [12], [15] and others, including dimensionality reduction for deep features.”
TreeTalk: Composition and Compression of Trees for Image Descriptions (2014, background): “Now that visual recognition algorithms are beginning to produce reliable estimates of image content (Perronnin et al., 2012; Deng et al., 2012a; Deng et al., 2010; Krizhevsky et al., 2012), the time seems ripe to begin exploring higher level semantic tasks.”
Towards Open World Recognition (2014, methodology): “Though more advanced features are available [30, 17, 37], extensive evaluation across features is beyond the scope of this work 2.”
Towards a Visual Turing Challenge (2014, background): “The human conceptualization divides these percepts into different instances, categories as well as sptio-temporal concepts.”
Label-Embedding for Image Classification (2015, positive): “Following [67], [44], we use a constant step size η t = η .”
Author manuscript, published in "International Journal of Computer Vision (2013)" International Journal of Computer Vision manuscript No. (will be inserted by the editor) Image Classification with the Fisher Vector: Theory and Practice (2013, unknown): “Given the size of these datasets, at each pass of the SGD routine we sample all positives but only a random subset of negatives (Perronnin et al, 2012, S ´ anchez and Perronnin, 2011).”
Averaged Least-Mean-Squares: Bias-Variance Trade-offs and Optimal Sampling Distributions (2015, background): “Then, it can be useful in practice to give more weight the less frequent class (see, e.g., Perronnin et al., 2012, and references therein).”
Label Partitioning For Sublinear Ranking (2013, background): “Wsabie was shown to be superior to unbalanced one-vs-rest (see also (Perronnin et al., 2012)), PAMIR (Grangier & Bengio, 2008), and k-nearest neighbors.”
Studying Relationships between Human Gaze, Description, and Computer Vision (2013, background): “Recent advances have started to look at problems of recognition at a human scale, classifying or localizing thousands of object categories with reasonable accuracy [19, 24, 5, 6, 18].”
From Large Scale Image Categorization to Entry-Level Categories (2013, background): “Although far from solved, algorithms have now advanced to the point where they can recognize or localize thousands of object categories with reasonable accuracy [17, 4, 3, 12].”
Distance-Based Image Classification: Generalizing to New Classes at Near-Zero Cost (2013, methodology): “9 ﬂat top-1 error [44].”
Deep Fisher Networks for Large-Scale Image Classification (2013, methodology): “Our SIFT feature extraction follows that of [21].”
Contextualizing Object Detection and Classification (2015, methodology): “For object classification, the prevalent methods [45], [31] use global features and discriminative modeling to achieve the goal.”
Efficient On-the-fly Category Retrieval Using ConvNets and GPUs (2014, unknown): “In particular, very high-dimensional feature vectors were required for state-of-the-art classification performance [12], [13], [14], but this incurred both a severe memory penalty (as features for the dataset need to be in memory for fast retrieval) and also a severe speed penalty (as computing a sc”
Consistency versus Realizable H-Consistency for Multiclass Classification (2013, background): “(We note that this is a standard practice for large-scale multiclass image classification, see e.g. (Weston et al., 2010; Lin et al., 2011; Perronnin et al., 2012)).”