Locally Scale-Invariant Convolutional Neural Networks

By Angjoo Kanazawa, Abhishek Sharma, David W. Jacobs. 2014

139 citations recorded by the source corpus.

Convolutional Neural Networks (ConvNets) have shown excellent results on many visual classification tasks. With the exception of ImageNet, these datasets are carefully crafted such that objects are well-aligned at similar scales. Naturally, the feature learning problem gets more challenging as the amount of variation in the data increases, as the models have to learn to be invariant to certain changes in appearance. Recent results on the ImageNet dataset show that given enough data, ConvNets can learn such invariances producing very discriminative features [1]. But could we do more: use less parameters, less data, learn more discriminative features, if certain invariances were built into the learning process? In this paper we present a simple model that allows ConvNets to learn features in a locally scale-invariant manner without increasing the number of model parameters. We show on a modified MNIST dataset that when faced with scale variation, building in scale-invariance allows ConvNets to learn more discriminative features with reduced chances of over-fitting.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Model Architectures

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 Model Architectures

Citations in the corpus

Making Convolutional Networks Shift-Invariant Again (2019, background): “Recent work has strived to build in additional types of in-variances, such as rotation, reﬂection, and scaling (Sifre & Mallat, 2013; Bruna & Mallat, 2013; Esteves et al., 2018; Kanazawa et al., 2014; Worrall et al., 2017; Cohen & Welling, 2016).”
Spatial Transformer Networks (2015, background): “In this section we discuss the prior work related to the paper, covering the central ideas of modelling transformations with neural networks [12, 13, 27], learning and analysing transformation-invariant representations [3, 5, 8, 17, 19, 25], as well as attention and detection mechanisms for feature ”
Deformable Convolutional Networks (2017, background): “Some works are devoted for speciﬁc transformations such as symmetry [13, 9], scale [29], and rotation [53].”
Learning Steerable Filters for Rotation Equivariant CNNs (2017, background): “Invariance to specific transformations can be achieved by applying them to the input and subsequently pooling their responses [11, 12, 13].”
Interaction-And-Aggregation Network for Person Re-Identification (2019, background): “Some works learn invariant CNN representations with respect to speciﬁc transformations such as symmetry [11], scale [21] and rotation [52].”
Under review as a conference paper at ICLR 2020 completely free-form sampling in image space : Spatial Transformers (2019, background): “However, this usually requires designs specific to certain kinds of deformation, such as shift, rotation, reflection and scaling (Sifre & Mallat, 2013; Bruna & Mallat, 2013; Kanazawa et al., 2016; Cohen & Welling, 2016; Worrall et al., 2017; Esteves et al., 2018).”
Robust Image Completion via Deep Feature Transformations (2019, methodology): “The authors in [52] proposed a method for creating scale-invariant CNNs, in which locally scale-invariant representations are obtained by applying filters at multiple scales and locations followed by max-pooling.”
Deep Learning for Generic Object Detection: A Survey (2018, unknown): “そのため,頑健性向上のために多くのアプローチが提示され,スケール [131, 21],回転 [21, 42, 284, 323],またはその両方 [126] など,様々な種類の変換に対して不変な CNN 表現の学習が目指されてきた.”
MAKING CONVOLUTIONAL NETWORKS SHIFT-INVARIANT AGAIN (2018, background): “Recent work has strived to build in additional types of invariances, such as rotation, reflection, and scaling (Sifre & Mallat, 2013; Bruna & Mallat, 2013; Esteves et al., 2017; Kanazawa et al., 2014; Worrall et al., 2017; Cohen & Welling, 2016).”
Dreaming More Data: Class-dependent Distributions over Diffeomorphisms for Learned Data Augmentation (2015, background): “For example, scale invariance is approximated via image-pyramid representations [17, 28], assuming a known set of scales which the classiﬁer must be invariant to.”
Deformable Kernels: Adapting Effective Receptive Fields for Object Deformation (2019, background): “However, this usually requires designs speciﬁc to certain kinds of deformation, such as shift, rotation, reﬂection and scaling (Sifre & Mallat, 2013; Bruna & Mallat, 2013; Kanazawa et al., 2016; Cohen & Welling, 2016; Worrall et al., 2017; Esteves et al., 2018).”
Scale equivariance in CNNs with vector fields (2018, methodology): “In this contribution, we propose to use a vector ﬁeld representation similar to the one in (Marcos et al., 2017) to obtain a locally scale-equivariant deep CNN.”
Under review as a conference paper at ICLR 2020 completely free-form sampling in image space : Spatial Transformers (2019, background): “However, this usually requires designs specific to certain kinds of deformation, such as shift, rotation, reflection and scaling (Sifre & Mallat, 2013; Bruna & Mallat, 2013; Kanazawa et al., 2016; Cohen & Welling, 2016; Worrall et al., 2017; Esteves et al., 2018).”
Universal Correspondence Network (2016, background): “However, explicitly handling such invariances in forms of data augmentation or explicit network structure yields higher accuracy in many tasks [17, 15, 13].”
Scale-Equivariant Steerable Networks (2019, methodology): “In SI-ConvNet by Kanazawa et al. (2014) the original network is modiﬁed such that, in each layer, the input is ﬁrst rescaled, then convolved and rescaled back to the original size.”