This is a PyTorch implementation of PCANet. Details are described in the original paper.
Unlike other implementations, the number of stages in PCANet can be set arbitrarily, rather than two. So the structure is more flexible.
- Python 3.5
- PyTorch==1.0.0
- sklearn, tensorboardX, numpy
python train.py
python eval.py --pretrained_path <path to trained PCANet model and SVM>
convolution kernel in stage 0
convolution kernel in stage 1
feature maps of an image in stage 0
feature maps of an image in stage 1
the accuracy rate in total testing data is 93.42%
convolution kernel in stage 0
convolution kernel in stage 1
feature maps of an image in stage 0
feature maps of an image in stage 1
the accuracy rate in total testing data is 93.48%
in progress
Chan T H , Jia K , Gao S , et al. PCANet: A Simple Deep Learning Baseline for Image Classification?[J]. IEEE Transactions on Image Processing, 2015, 24(12):5017-5032.