Learning to Compare Image Patches via Convolutional Neural Networks

Abstract

In this paper we show how to learn directly from image data (i.e., without resorting to manually-designed features) a general similarity function for comparing image patches, which is a task of fundamental importance for many computer vision problems. To encode such a function, we opt for a CNN-based model that is trained to account for a wide variety of changes in image appearance. To that end, we explore and study multiple neural network architectures, which are specifically adapted to this task. We show that such an approach can significantly outperform the state-of-the-art on several problems and benchmark datasets.

CVPR 2015

References in corpus (2)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Descriptor Matching with Convolutional Neural Networks: a Comparison to SIFT

Cited by in corpus (26)

• End-to-End Learning of Geometry and Context for Deep Stereo Regression
• Generic 3D Representation via Pose Estimation and Matching
• CliqueCNN: Deep Unsupervised Exemplar Learning
• RankIQA: Learning from Rankings for No-reference Image Quality Assessment
• A Deep Metric for Multimodal Registration
• Localizing and Orienting Street Views Using Overhead Imagery
• Cooperative Learning with Visual Attributes
• FCSS: Fully Convolutional Self-Similarity for Dense Semantic Correspondence
• Local Feature Detectors, Descriptors, and Image Representations: A Survey
• Learning Spread-out Local Feature Descriptors
• A Performance Evaluation of Local Features for Image Based 3D Reconstruction
• Real-Time Dense Stereo Embedded in A UAV for Road Inspection
• Sunrise or Sunset: Selective Comparison Learning for Subtle Attribute Recognition
• Deep Multi-Modal Image Correspondence Learning
• Deep Stereo Matching with Dense CRF Priors
• FrameNet: Learning Local Canonical Frames of 3D Surfaces from a Single RGB Image
• Spot the Difference by Object Detection
• AutoScaler: Scale-Attention Networks for Visual Correspondence
• Siamese convolutional networks based on phonetic features for cognate identification
• Detect, Replace, Refine: Deep Structured Prediction For Pixel Wise Labeling
• Proposal Flow: Semantic Correspondences from Object Proposals
• Visual-Interactive Similarity Search for Complex Objects by Example of Soccer Player Analysis
• Image Patch Matching Using Convolutional Descriptors with Euclidean Distance
• Image Matching via Loopy RNN
• Automatic Synchronization of Multi-User Photo Galleries
• DCI: Discriminative and Contrast Invertible Descriptor