Convolutional Neural Networks learn compact local image descriptors

TL;DR

This paper demonstrates that a standard deep convolutional neural network, when combined with an appropriate loss function, can learn compact local image descriptors that match the performance of existing leading methods.

Contribution

It shows that a straightforward CNN architecture with the right loss function can effectively learn local image descriptors, simplifying the approach compared to prior complex methods.

Findings

CNNs can learn compact local descriptors

Performance comparable to state-of-the-art methods

Simplifies local descriptor learning process

Abstract

A standard deep convolutional neural network paired with a suitable loss function learns compact local image descriptors that perform comparably to state-of-the art approaches.