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 | 67 | 9.84% |
Object Detection | 46 | 6.75% |
Classification | 37 | 5.43% |
Quantization | 31 | 4.55% |
Semantic Segmentation | 29 | 4.26% |
Decoder | 12 | 1.76% |
Instance Segmentation | 10 | 1.47% |
Management | 9 | 1.32% |
Ensemble Learning | 8 | 1.17% |
Component | Type |
|
---|---|---|
🤖 No Components Found | You can add them if they exist; e.g. Mask R-CNN uses RoIAlign |