Revisiting Edge Detection in Convolutional Neural Networks
This work addresses the problem of robust edge detection for computer vision models, particularly concerning their reliance on color information which can lead to brittleness.
This paper demonstrates that edges are poorly represented in the initial layers of popular CNN architectures like VGG-16 and ResNet, which instead rely on color information. To address this, the authors propose edge-detection units that improve robustness against color noise and generate qualitatively different representations.
The ability to detect edges is a fundamental attribute necessary to truly capture visual concepts. In this paper, we prove that edges cannot be represented properly in the first convolutional layer of a neural network, and further show that they are poorly captured in popular neural network architectures such as VGG-16 and ResNet. The neural networks are found to rely on color information, which might vary in unexpected ways outside of the datasets used for their evaluation. To improve their robustness, we propose edge-detection units and show that they reduce performance loss and generate qualitatively different representations. By comparing various models, we show that the robustness of edge detection is an important factor contributing to the robustness of models against color noise.