This directory contains all of the ARM computational kernels used the paper "Fast Sparse ConvNets" submitted to CVPR 2020.
spmm-NxM-[scalar,neonfma].c - Sparse Matrix Multiplication with an unroll of N in the HW dimension and a block size of M in the channel output dimension.
dwconv-KxKsSpP-[scalar,neonfma].c - Depthwise Convolution with a filter of KxK, a stride of S and symmetric padding of P.
gavgpool-[scalar,neon]-xN.c - Global average pooling unrolled over N rows.
conv-3x3s2p1c3x4-[scalar,neonfma]-KxK - Full 3x3 stride 2 convolution with HWC input and CHW output. Operates on 3 input channels (only supported) and 4 output channels in the inner loop. Produces a KxK output.
The MobileNetV1 models have a block size of 4 in the last block, otherwise they are unstructured.
The MobileNetV2 models have a block size of 2 from block 11 onwards, otherwise they are unstructured. The exception is the width 1.8, 80% sparse model which unstructured throughout.
The first full convolution and the final fully connected layer are both dense in all models.
| Model | Top-1 Accuracy | Sparsity | Download |
|---|---|---|---|
| MobileNetV1 .75 | 64.4% | 90% | link |
| MobileNetV1 1.0 | 68.4% | 90% | link |
| MobileNetV1 1.4 | 72.0% | 90% | link |
| MobileNetV2 .8 | 65.2% | 85% | link |
| MobileNetV2 1.15 | 70.2% | 85% | link |
| MobileNetV2 1.4 | 72.0% | 85% | link |
| MobileNetV2 1.8 | 74.9% | 80% | link |
| MobileNetV2 2.0 | 74.5% | 85% | link |