A Simple Fix for Convolutional Neural Network via Coordinate Embedding

TL;DR

This paper introduces a simple coordinate embedding method to enhance CNNs by incorporating pixel coordinate information, improving their ability to handle affine transformations and increasing robustness.

Contribution

The paper proposes a straightforward coordinate embedding technique that can be applied to existing CNNs without architectural changes, boosting performance and robustness.

Findings

Significant performance improvements on traffic sign detection.

Enhanced robustness to affine transformations.

Easy integration with pre-trained models.

Abstract

Convolutional Neural Networks (CNN) has been widely applied in the realm of computer vision. However, given the fact that CNN models are translation invariant, they are not aware of the coordinate information of each pixel. Thus the generalization ability of CNN will be limited since the coordinate information is crucial for a model to learn affine transformations which directly operate on the coordinate of each pixel. In this project, we proposed a simple approach to incorporate the coordinate information to the CNN model through coordinate embedding. Our approach does not change the downstream model architecture and can be easily applied to the pre-trained models for the task like object detection. Our experiments on the German Traffic Sign Detection Benchmark show that our approach not only significantly improve the model performance but also have better robustness with respect to the affine transformation.