Rotated Ring, Radial and Depth Wise Separable Radial Convolutions
This work addresses the issue of rotation sensitivity in neural networks for computer vision applications, offering a novel method that is incremental in improving invariance without data augmentation.
The paper tackles the problem of deep neural networks losing accuracy under image rotations and proposes trainable rotation invariant convolutions to address this without increasing dataset size. The approach demonstrates rotational invariance across models and datasets, though it is more computationally intensive than standard convolutions.
Simple image rotations significantly reduce the accuracy of deep neural networks. Moreover, training with all possible rotations increases the data set, which also increases the training duration. In this work, we address trainable rotation invariant convolutions as well as the construction of nets, since fully connected layers can only be rotation invariant with a one-dimensional input. On the one hand, we show that our approach is rotationally invariant for different models and on different public data sets. We also discuss the influence of purely rotational invariant features on accuracy. The rotationally adaptive convolution models presented in this work are more computationally intensive than normal convolution models. Therefore, we also present a depth wise separable approach with radial convolution. Link to CUDA code https://atreus.informatik.uni-tuebingen.de/seafile/d/8e2ab8c3fdd444e1a135/