A Deep Learning Model by combining Covolutional an Involutional Neural Networks.
In-Conv Model is an attempt to combine Involution and Convolutional Neural Network.
Involutional Kerenl is both location specific and channel-agnostic. Convolution, on the other hand is location-agnostic and channel specific.
In-Conv model combines these orthagonal properties to experiment with the CIFAR10 dataset.
Feel free to experiment.
Credits/Sources: I Involution: Inverting the Inherence of Convolution for VisualRecognition - https://arxiv.org/abs/2103.06255 Involution Implementation : https://github.com/keras-team/keras-io/blob/master/examples/vision/involution.py