Skip to Main content Skip to Navigation
Journal articles

Rotation Invariant Networks for Image Classification for HPC and Embedded Systems

Abstract : Convolutional Neural Network (CNNs) models’ size reduction has recently gained interest due to several advantages: energy cost reduction, embedded devices, and multi-core interfaces. One possible way to achieve model reduction is the usage of Rotation-invariant Convolutional Neural Networks because of the possibility of avoiding data augmentation techniques. In this work, we present the next step to obtain a general solution to endowing CNN architectures with the capability of classifying rotated objects and predicting the rotation angle without data-augmentation techniques. The principle consists of the concatenation of a representation mapping transforming rotation to translation and a shared weights predictor. This solution has the advantage of admitting different combinations of various basic, existing blocks. We present results obtained using a Gabor-filter bank and a ResNet feature backbone compared to previous other solutions. We also present the possibility to select between parallelizing the network in several threads for energy-aware High Performance Computing (HPC) applications or reducing the memory footprint for embedded systems. We obtain a competitive error rate on classifying rotated MNIST and outperform existing state-of-the-art results on CIFAR-10 when trained on up-right examples and validated on random orientations.
Complete list of metadata
Contributor : Petr Dokladal Connect in order to contact the contributor
Submitted on : Tuesday, January 12, 2021 - 9:12:58 AM
Last modification on : Friday, April 1, 2022 - 3:45:00 AM


Distributed under a Creative Commons Attribution 4.0 International License

Links full text



Rosemberg Rodriguez Salas, Petr Dokládal, Eva Dokladalova. Rotation Invariant Networks for Image Classification for HPC and Embedded Systems. Electronics, Penton Publishing Inc., 2021, 10 (2), pp.139. ⟨10.3390/electronics10020139⟩. ⟨hal-03106734⟩



Record views