Random Padding Data Augmentation
This addresses a domain-specific issue in CNN-based image recognition by mitigating position learning interference, though it is incremental as it builds on existing padding and augmentation techniques.
The paper tackled the problem that CNNs learning position information hinders learning feature relationships, and introduced Random Padding, a parameter-free method that adds zero-padding randomly to half of feature map borders, consistently improving image classification performance over strong baselines.
The convolutional neural network (CNN) learns the same object in different positions in images, which can improve the recognition accuracy of the model. An implication of this is that CNN may know where the object is. The usefulness of the features' spatial information in CNNs has not been well investigated. In this paper, we found that the model's learning of features' position information hindered the learning of the features' relationship. Therefore, we introduced Random Padding, a new type of padding method for training CNNs that impairs the architecture's capacity to learn position information by adding zero-padding randomly to half of the border of feature maps. Random Padding is parameter-free, simple to construct, and compatible with the majority of CNN-based recognition models. This technique is also complementary to data augmentations such as random cropping, rotation, flipping and erasing, and consistently improves the performance of image classification over strong baselines.