Pointwise Convolution

Pointwise Convolution is a type of convolution that uses a 1x1 kernel: a kernel that iterates through every single point. This kernel has a depth of however many channels the input image has. It can be used in conjunction with depthwise convolutions to produce an efficient class of convolutions known as depthwise-separable convolutions.

Image Credit: Chi-Feng Wang


Paper Code Results Date Stars


Task Papers Share
Image Classification 76 11.41%
Object Detection 51 7.66%
Classification 38 5.71%
Quantization 34 5.11%
Semantic Segmentation 34 5.11%
Instance Segmentation 11 1.65%
Model Compression 9 1.35%
Benchmarking 9 1.35%
Edge-computing 8 1.20%


Component Type
🤖 No Components Found You can add them if they exist; e.g. Mask R-CNN uses RoIAlign