Training ResNet18 or AlexNet for classification

By Linfeng

caffe

Please build caffe

1.Make Dataset for classification

Data
#train
##face
##smoking
#val
##face
##smoking

2.Convert Dataset to lmdb

cd Data python generate_txt.py #get train.txt and val.txt

cd .. ./create_imagenet.sh #get lmdb Dataset

3.Get mean.binaryproto and mean.npy

sudo ~/PATH-TO-CAFFE/caffe/build/tools/compute_image_mean ./Data/train_lmdb #get the mean.binaryproto

sudo python convert_mean.py mean.binaryproto mean.npy #get mean.npy

4.training

Please modfiy the path of your dataset and hyperparameters of solver.prototxt run train.sh for ResNet-18 run train_alex.sh for Alex-Net