# plt.rcParams['figure.figsize'] = (15.0, 8.0) # set default size of plots
# plt.rcParams['image.interpolation'] = 'nearest'
# plt.rcParams['image.cmap'] = 'gray'
# # for auto-reloading extenrnal modules
# # see http://stackoverflow.com/questions/1907993/autoreload-of-modules-in-ipython
# %load_ext autoreload
# %autoreload 2
# set up fransform parameters to find the bounding box, crop and resize the images
# transfrom parameters
pad_inf = 1 # infimum scale
pad_sup = 1.5 # supremum scale
size = 100 # resize size
shift = 3 # radomly shift the bounding box
lcn = True # local contrast normalization
trans = Transform(pad_inf, pad_sup, size=size, shift=shift, lcn=lcn)
# set up the model and trainer for training
data_dir = 'data/FLIC-full' # store image
data_info_file = '%s/train_joints.csv' % data_dir # store corrdinates of joints for training data
model = LinearNet_flic(size)
trainer = NetworkTrainer(data_dir=data_dir, data_info_file=data_info_file,
model=model, trans=trans)
if __name__ == '__main__':
best_model, train_loss_history, val_loss_history = trainer.train_bgd()
help='mode sgd or bgd')
args = parser.parse_args()
# set up fransform parameters to find bounding box, crop and resize images
# transfrom parameters
pad_inf = 1.5 # infimum scale
pad_sup = 2.0 # supremum scale
size = 100 # resize size
shift = 3 # radomly shift the bounding box
lcn = True # local contrast normalization
trans = Transform(pad_inf, pad_sup, size=size, shift=shift, lcn=lcn)
# set up the model and trainer for training
data_dir = 'data/FLIC-full' # store image
# store corrdinates of joints for training datas
data_info_file = '%s/train_joints.csv' % data_dir
model = LinearNet_flic(size)
trainer = NetworkTrainer(data_dir=data_dir, data_info_file=data_info_file,
model=model, num_epochs=100, trans=trans)
if args.mode == 'sgd':
results = trainer.train()
results_file = 'results/sgd_trained_resutls.chainer'
elif args.mode == 'bgd':
results = trainer.train_bgd()
results_file = 'results/bgd_trained_resutls.chainer'
else:
raise Exception('Unrecognized mode type "%s"' % args.mode)
# best_model, train_loss_batch_history,
# train_loss_epoch_history, val_loss_epoch_history
with open(results_file, 'wb', pickle.HIGHEST_PROTOCOL) as pickle_file:
pickle.dump(results, pickle_file)