def test():
# Set input and output dirs
input_dir = "/storage/cfmata/deeplab/crf_rnn/crfasrnn_keras/data/horse_fine/images_orig/"
output_dir = "/storage/cfmata/deeplab/crf_rnn/crfasrnn_keras/image_results/horse_fine/fcn/"
input_size = 224
num_crf_iter = 10
saved_model_path = 'results/horse_fine/horse_fine_weights.500-0.53'
#model = get_crfrnn_model_def()
model = load_model_gby('fcn_RESNET50_8s', input_size, 22, num_crf_iter)
model.load_weights(saved_model_path)
im_list = open("lst/horsecoarse_test.txt").readlines()
im_list = [f[:-1] for f in im_list]
for img in im_list:
img_data, img_h, img_w = util.get_preprocessed_image(input_dir + img +
".jpg")
probs = model.predict(img_data, verbose=False,
batch_size=1)[0, :, :, :]
segmentation = util.get_label_image(probs, img_h, img_w)
print(output_dir + img)
segmentation.save(output_dir + img[:-4] + ".png")
#
# print("batch sizes train ", batch_sizes_train)
# print("batch sizes val ", batch_sizes_val)
# print("batch sizes total ", batch_sizes_total)
# pdb.set_trace()
# with tf.device('/cpu:0'):
# input_image, output_image = data_augmentation(input_image, output_image)
# ===============
# LOAD model:
# ===============
# for training:
num_crf_iterations = 5
model = load_model_gby(args.model, INPUT_SIZE, nb_classes, num_crf_iterations, args.finetune_path, batch_size)
#,batch_sizes_train, batch_sizes_val, batch_sizes_total)
# if resuming training:
if (args.weights is not None) and (os.path.exists(args.weights)):
print("loading weights %s.."% args.weights)
model.load_weights(args.weights)
model.summary()
print('trining model %s..'% model.name)
# ===============
# LOAD sp segment:
# ===============
# if model.sp_flag:
# segments_train = load_segmentations(ds.segments_dir, ds.train_list, INPUT_SIZE)
# ===============
# LOAD model:
# ===============
model_name = args.model
model_path_name = args.weights
base_img_name = []
print('====================================================================================')
print(model_path_name)
print('====================================================================================')
finetune_path = ''
#pdb.set_trace()
model = load_model_gby(model_name, INPUT_SIZE, n_classes, num_crf_iterations, finetune_path)
#loading weights:
model.load_weights(model_path_name)
# Computing prediction:
# ------------------------------
print('computing prediction..')
if args.folderpath==None:
img_org = cv2.imread(args.imagepath)
x = load_image(img_org, INPUT_SIZE)
base_img_name.append(os.path.splitext(os.path.basename(args.imagepath))[0])
else:
X = []
outdirName = args.folderpath+'out/'
ds = load_dataset(args.dataset, INPUT_SIZE)
print(ds.X_train.shape, ds.y_train.shape)
print(ds.X_test.shape, ds.y_test.shape)
nb_classes = ds.nb_classes
# pdb.set_trace()
# with tf.device('/cpu:0'):
# input_image, output_image = data_augmentation(input_image, output_image)
# ===============
# LOAD model:
# ===============
# for training:
num_crf_iterations = 5
model = load_model_gby(args.model, INPUT_SIZE, nb_classes,
num_crf_iterations)
# if resuming training:
if (args.weights is not None) and (os.path.exists(args.weights)):
print("loading weights %s.." % args.weights)
model.load_weights(args.weights)
model.summary()
print('training model %s..' % model.name)
# ===============
# LOAD sp segment:
# ===============
if model.sp_flag:
segments_train = load_segmentations(ds.segments_dir, ds.train_list,
INPUT_SIZE)