As known, modern most popular CNN (convolutional neural network): VGG/ResNet (FasterRCNN), SSD, Yolo, Yolo v2, DenseBox, DetectNet - are not rotate invariant: __Are modern CNN (convolutional neural network) as DetectNet rotate invariant?__

Also known, that there are several neural networks with rotate-invariance object detection:

Rotation-Invariant Neoperceptron 2006 (

__PDF__):__https://www.researchgate.net/publication/224649475_Rotation-Invariant_Neoperceptron__Learning rotation invariant convolutional filters for texture classification 2016 (

__PDF__):__https://arxiv.org/abs/1604.06720__RIFD-CNN: Rotation-Invariant and Fisher Discriminative Convolutional Neural Networks for Object Detection 2016 (

__PDF__):__http://www.cv-foundation.org/openaccess/content_cvpr_2016/html/Cheng_RIFD-CNN_Rotation-Invariant_and_CVPR_2016_paper.html__Encoded Invariance in Convolutional Neural Networks 2014 (

__PDF__)Rotation-invariant convolutional neural networks for galaxy morphology prediction (

__PDF__):__https://arxiv.org/abs/1503.07077__Learning Rotation-Invariant Convolutional Neural Networks for Object Detection in VHR Optical Remote Sensing Images 2016:

__http://ieeexplore.ieee.org/document/7560644/__

We know, that in such image-detection competitions as: IMAGE-NET, MSCOCO, PASCAL VOC - used networks ensembles (simultaneously some neural networks). Or networks ensembles in single net such as ResNet (__Residual Networks Behave Like Ensembles of Relatively Shallow Networks__)

But are used rotation invariant network ensembles in winners like as MSRA, and if not, then why? Why in ensemble the additional rotation-invariant network does not add accuracy to detect certain objects such as aircraft objects - which images is done at a different angles of rotation?

It can be:

aircraft objects which are photographed from the ground

or ground objects which are photographed from the air

Why rotation-invariant neural networks are not used in winners of the popular object-detection competitions?