Rotation invariant CNN using scattering transform for image classification
This work addresses rotation invariance in image classification, which is crucial for applications like automatic re-orientation of datasets, but it appears incremental as it builds on existing scattering transform networks.
The paper tackles the problem of deep convolutional neural networks being sensitive to input rotations by proposing a rotation-invariant CNN using scattering transform, which can predict angular orientation without angle-annotated data and map rotations continuously to a circular space, validated with upright and randomly rotated samples.
Deep convolutional neural networks accuracy is heavily impacted by rotations of the input data. In this paper, we propose a convolutional predictor that is invariant to rotations in the input. This architecture is capable of predicting the angular orientation without angle-annotated data. Furthermore, the predictor maps continuously the random rotation of the input to a circular space of the prediction. For this purpose, we use the roto-translation properties existing in the Scattering Transform Networks with a series of 3D Convolutions. We validate the results by training with upright and randomly rotated samples. This allows further applications of this work on fields like automatic re-orientation of randomly oriented datasets.