PocketNet is a face recognition model family discovered through neural architecture search. The training is based on multi-step knowledge distillation.
Source: PocketNet: Extreme Lightweight Face Recognition Network using Neural Architecture Search and Multi-Step Knowledge DistillationPaper | Code | Results | Date | Stars |
---|
Component | Type |
|
---|---|---|
Batch Normalization
|
Normalization | |
Convolution
|
Convolutions | |
Depthwise Convolution
|
Convolutions | |
Depthwise Separable Convolution
|
Convolutions | |
PReLU
|
Activation Functions |