MMALIB User Guide
Convolutional Neural Networks (CNN) kernels

Introduction

This module consists of kernels to implement the core computations occurring in the context of convolutional neural networks.

Sub Modules

 MMALIB_CNN_convolveBias_row_ixX_ixX_oxX
 Kernel for computing dense CNN convolution with row based processing.
 
 MMALIB_CNN_convolve_col_smallNo_highPrecision
 NOTE: This API is now a wrapper to MMALIB_CNN_convolve_col_smallNo_highPrecision_pointwisePost with the lutValues input argument set to NULL. It is recommended to call MMALIB_CNN_convolve_col_smallNo_highPrecision_pointwisePost directly.
 
 MMALIB_CNN_convolve_col_smallNo_highPrecision_pointwisePost
 Kernel for computing CNN-style 2D convolution using column major data ordering on the input and output feature maps. This approach computes more quickly if filter grouping is chosen such that Ni=No=1, or if filter grouping is chosen such that NiFrFc < MMA_SIZE, otherwise use regular convolution method MMALIB_CNN_convolve_row_ixX_ixX_oxX. This kernel is also referred to as depth-wise convolution.
 
 MMALIB_CNN_convolve_col_smallNo_ixX_ixX_oxX
 Kernel for computing CNN-style 2D convolution using column major data ordering on the input and output feature maps. This approach computes more quickly if filter grouping is chosen such that Ni=No=1, or if filter grouping is chosen such that NiFrFc < MMA_SIZE, otherwise use regular convolution method MMALIB_CNN_convolve_row_ixX_ixX_oxX. This kernel is also referred to as depth-wise convolution.
 
 MMALIB_CNN_deconvolve_row_ixX_ixX_oxX
 Kernel for computing dense CNN deconvolution with row-based processing and matrix-matrix multiplication.
 
 MMALIB_CNN_fullyConnectedBias_ixX_ixX_oxX
 Kernel provides compute functionality of Fully Connected Layer: \( Y^T = X^T \times H^T + B^T\).
 
 MMALIB_CNN_tensor_convert_ixX_oxX
 Kernel for converting tensors of various datatypes and formats.