Scale-Invariant Convolutional Neural Networks

Abstract

Even though convolutional neural networks (CNN) has achieved near-human performance in various computer vision tasks, its ability to tolerate scale variations is limited. The popular practise is making the model bigger first, and then train it with data augmentation using extensive scale-jittering. In this paper, we propose a scaleinvariant convolutional neural network (SiCNN), a modeldesigned to incorporate multi-scale feature exaction and classification into the network structure. SiCNN uses a multi-column architecture, with each column focusing on a particular scale. Unlike previous multi-column strategies, these columns share the same set of filter parameters by a scale transformation among them. This design deals with scale variation without blowing up the model size. Experimental results show that SiCNN detects features at various scales, and the classification result exhibits strong robustness against object scale variations.

This paper is submitted for CVPR2015

References in corpus (6)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Improving neural networks by preventing co-adaptation of feature detectors
• Practical Bayesian Optimization of Machine Learning Algorithms
• Going Deeper with Convolutions
• Visualizing and Understanding Convolutional Networks
• Multi-scale Orderless Pooling of Deep Convolutional Activation Features

Cited by in corpus (33)

• Deep High-Resolution Representation Learning for Visual Recognition
• Why do deep convolutional networks generalize so poorly to small image transformations?
• Scale-aware Fast R-CNN for Pedestrian Detection
• Deep High-Resolution Representation Learning for Human Pose Estimation
• Scale-Equivariant Steerable Networks
• Scale equivariance in CNNs with vector fields
• Why Self-Attention? A Targeted Evaluation of Neural Machine Translation Architectures
• Neural Networks Enforcing Physical Symmetries in Nonlinear Dynamical Lattices: The Case Example of the Ablowitz-Ladik Model
• Stacked Pooling: Improving Crowd Counting by Boosting Scale Invariance
• Scale-invariant scale-channel networks: Deep networks that generalise to previously unseen scales
• Particle Identification In Camera Image Sensors Using Computer Vision
• Investigation to answer three key questions concerning plant pest identification and development of a practical identification framework
• Scaling-Translation-Equivariant Networks with Decomposed Convolutional Filters
• Learning Online Visual Invariances for Novel Objects via Supervised and Self-Supervised Training
• Towards Learning Affine-Invariant Representations via Data-Efficient CNNs
• Temporal Transformer Networks: Joint Learning of Invariant and Discriminative Time Warping
• Downbeat Tracking with Tempo-Invariant Convolutional Neural Networks
• Dynamic Stale Synchronous Parallel Distributed Training for Deep Learning
• Non-Parametric Transformation Networks
• RAIN: A Simple Approach for Robust and Accurate Image Classification Networks
• DCT-Conv: Coding filters in convolutional networks with Discrete Cosine Transform
• Learning Visual Representations for Transfer Learning by Suppressing Texture
• Learning to Point and Count
• Saliency Preservation in Low-Resolution Grayscale Images
• Learning Translation Invariance in CNNs
• Rotation-Invariant Autoencoders for Signals on Spheres
• Deformation Robust Roto-Scale-Translation Equivariant CNNs
• Scale-wise Convolution for Image Restoration
• Do CNNs Encode Data Augmentations?
• Learning scale-variant and scale-invariant features for deep image classification
• REST: Performance Improvement of a Black Box Model via RL-based Spatial Transformation
• A Progressively-trained Scale-invariant and Boundary-aware Deep Neural Network for the Automatic 3D Segmentation of Lung Lesions
• Feature Lenses: Plug-and-play Neural Modules for Transformation-Invariant Visual Representations