Mind the Pad -- CNNs can Develop Blind Spots
arXiv:2010.02178
Abstract
We show how feature maps in convolutional networks are susceptible to spatial bias. Due to a combination of architectural choices, the activation at certain locations is systematically elevated or weakened. The major source of this bias is the padding mechanism. Depending on several aspects of convolution arithmetic, this mechanism can apply the padding unevenly, leading to asymmetries in the learned weights. We demonstrate how such bias can be detrimental to certain tasks such as small object detection: the activation is suppressed if the stimulus lies in the impacted area, leading to blind spots and misdetection. We propose solutions to mitigate spatial bias and demonstrate how they can improve model accuracy.
Appendix E available at https://drive.google.com/file/d/1bIvRQJIBwJbKTfpg0hNaFX2ThuuDO8PU/view?usp=sharing
References in corpus (10)
- PyTorch: An Imperative Style, High-Performance Deep Learning Library
- MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
- TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems
- Understanding the Effective Receptive Field in Deep Convolutional Neural Networks
- Making Convolutional Networks Shift-Invariant Again
- Inverting Gradients -- How easy is it to break privacy in federated learning?
- How Much Position Information Do Convolutional Neural Networks Encode?
- Object Detection Through Exploration With A Foveated Visual Field
- Partial Convolution based Padding
- Translation Insensitive CNNs