Quantifying Translation-Invariance in Convolutional Neural Networks

Abstract

A fundamental problem in object recognition is the development of image representations that are invariant to common transformations such as translation, rotation, and small deformations. There are multiple hypotheses regarding the source of translation invariance in CNNs. One idea is that translation invariance is due to the increasing receptive field size of neurons in successive convolution layers. Another possibility is that invariance is due to the pooling operation. We develop a simple a tool, the translation-sensitivity map, which we use to visualize and quantify the translation-invariance of various architectures. We obtain the surprising result that architectural choices such as the number of pooling layers and the convolution filter size have only a secondary effect on the translation-invariance of a network. Our analysis identifies training data augmentation as the most important factor in obtaining translation-invariant representations of images using convolutional neural networks.

References in corpus (2)

• Adam: A Method for Stochastic Optimization
• Unsupervised Learning of Invariant Representations in Hierarchical Architectures

Cited by in corpus (18)

• Why do deep convolutional networks generalize so poorly to small image transformations?
• CNN-based Methods for Object Recognition with High-Resolution Tactile Sensors
• 2022 Review of Data-Driven Plasma Science
• An Image is Worth 16x16 Words, What is a Video Worth?
• Evading Defenses to Transferable Adversarial Examples by Translation-Invariant Attacks
• Computational Separation Between Convolutional and Fully-Connected Networks
• Learning Online Visual Invariances for Novel Objects via Supervised and Self-Supervised Training
• Deep Scattering Network with Max-pooling
• Comparison of Methods Generalizing Max- and Average-Pooling
• Reconstructing Neutrino Energy using CNNs for GeV Scale IceCube Events
• The Effect of Learning Strategy versus Inherent Architecture Properties on the Ability of Convolutional Neural Networks to Develop Transformation Invariance
• Anti-aliasing Deep Image Classifiers using Novel Depth Adaptive Blurring and Activation Function
• Using Convolutional Neural Networks to Reconstruct Energy of GeV Scale IceCube Neutrinos
• Unsupervised Learning of Lagrangian Dynamics from Images for Prediction and Control
• Invariance Measures for Neural Networks
• RAIN: A Simple Approach for Robust and Accurate Image Classification Networks
• Learning Translation Invariance in CNNs
• Grounding inductive biases in natural images:invariance stems from variations in data