HetConv: Heterogeneous Kernel-Based Convolutions for Deep CNNs
This addresses efficiency issues in deep learning for practitioners needing faster CNNs, though it is incremental as it builds on existing architectures.
The paper tackles the problem of high computational cost in deep CNNs by proposing HetConv, a convolution operation using heterogeneous kernels, which reduces FLOPs by 3X to 8X while maintaining or improving accuracy on architectures like VGG and ResNet.
We present a novel deep learning architecture in which the convolution operation leverages heterogeneous kernels. The proposed HetConv (Heterogeneous Kernel-Based Convolution) reduces the computation (FLOPs) and the number of parameters as compared to standard convolution operation while still maintaining representational efficiency. To show the effectiveness of our proposed convolution, we present extensive experimental results on the standard convolutional neural network (CNN) architectures such as VGG \cite{vgg2014very} and ResNet \cite{resnet}. We find that after replacing the standard convolutional filters in these architectures with our proposed HetConv filters, we achieve 3X to 8X FLOPs based improvement in speed while still maintaining (and sometimes improving) the accuracy. We also compare our proposed convolutions with group/depth wise convolutions and show that it achieves more FLOPs reduction with significantly higher accuracy.