raw_ds = [np.expand_dims(pygt.normalize(np.array(Image.open(raw_path[i]).convert('L'), 'f')),0) for i in range(0, num_images)]
datasets = []
for i in range(0,len(raw_ds)):
dataset = {}
dataset['data'] = raw_ds[i]
datasets += [dataset]
test_net_file = 'net.prototxt'
test_device = 0
pygt.caffe.set_devices((test_device,))
caffemodels = pygt.get_caffe_models('net')
test_net = pygt.init_testnet(test_net_file, trained_model=caffemodels[-1][1], test_device=test_device)
def process_data_slice_callback(input_specs, batch_size, dataset_indexes, offsets, dataset_combined_sizes, data_arrays, slices):
# Nothing to process here
pass
output_arrays = []
pygt.process(test_net, datasets, ['aff_pred', 'smax_pred'], output_arrays, process_data_slice_callback)
for i in range(0, len(output_arrays)):
for key in output_arrays[i].keys():
outhdf5 = h5py.File('output/' + key + str(i) + '.h5', 'w')
outdset = outhdf5.create_dataset('main', np.shape(output_arrays[i][key]), np.float32, data=output_arrays[i][key])
outhdf5.close()
raw_ds = [np.array(Image.open(raw_path[i]).convert('L'), 'f') for i in range(0, num_images)]
datasets = []
test_datasets = []
for i in range(0,1):
dataset = {}
dataset['data'] = np.expand_dims(pygt.normalize(np.asarray(raw_ds, float32)), 0)
datasets += [dataset]
test_net_file = 'net.prototxt'
test_device = 0
pygt.caffe.set_devices((test_device,))
caffemodels = pygt.get_caffe_models('net')
test_net = pygt.init_testnet(test_net_file, trained_model=caffemodels[-1][1], test_device=test_device)
def process_data_slice_callback(input_specs, batch_size, dataset_indexes, offsets, dataset_combined_sizes, data_arrays, slices):
# Nothing to process here
pass
output_arrays = []
pygt.process(test_net, datasets, ['prob'], output_arrays, process_data_slice_callback)
for i in range(0, len(output_arrays)):
for key in output_arrays[i].keys():
outhdf5 = h5py.File('output/' + key + str(i) + '.h5', 'w')
outdset = outhdf5.create_dataset('main', np.shape(output_arrays[i][key]), np.float32, data=output_arrays[i][key])
outhdf5.close()
