Generate random image buffers, instantiate hybrid alexnet-overfeat architecture, then begin training
do the following: ./main_random.sh solver_alexoverfeat.prototoxt
If the train, val, and random_train_leveldb, random_val_leveldb folders already exist, you can do: ./run_training.sh solver_alexoverfeat.prototxt
To change the input dimensions of the images: Open GenerateImages.m and change the first two lines note: if train, val, and random_train_leveldb, random_val_leveldb folders already exist, they will need to be deleted before main_random can be run again
Note: this is the same exact procedure for using the second version: randomNewCaffe-LMDB/ This version works for the latest version of caffe (as of September 2014), and uses the LMDB file format of caffe as opposed to the leveldb version.