Color Object Classification using Multi-channel Zernike Moments-Based Rotation Invariant Bag-of-Visual-Words and Deep Convolutional Neural Networks

•Rotation invariant color and shape descriptor is proposed using bag-of-visual-words.•Feature maps of CNNs are used to derive invariant features.•Several feature maps are pooled to form rotation invariant descriptor.•Multi-channel Zernike moments are used for pooling feature maps.•High recognition rates are achieved surpassing the state-of-the-art CNNs.