def main(_): if not os.path.exists(args.checkpoint_dir): os.makedirs(args.checkpoint_dir) if not os.path.exists(args.sample_dir): os.makedirs(args.sample_dir) # os.makedirs(args.sample_dir +"/A") # os.makedirs(args.sample_dir +"/B") ### 因為寫在function裡會一直被呼叫到,所以我才拉出來main寫喔! # os.makedirs(args.sample_dir +"/to_curved/big") # os.makedirs(args.sample_dir +"/to_curved/big-left-top") # os.makedirs(args.sample_dir +"/to_curved/small-seen") # os.makedirs(args.sample_dir +"/to_curved/small-unseen") # os.makedirs(args.sample_dir +"/to_straight/big") # os.makedirs(args.sample_dir +"/to_straight/big-left-top") # os.makedirs(args.sample_dir +"/to_straight/small-seen") # os.makedirs(args.sample_dir +"/to_straight/small-unseen") ### 因為寫在function裡會一直被呼叫到,所以我才拉出來main寫喔! os.makedirs(args.sample_dir + "/to_straight/crop-accurate") if not os.path.exists(args.test_dir): os.makedirs(args.test_dir) tfconfig = tf.ConfigProto(allow_soft_placement=True) tfconfig.gpu_options.allow_growth = True with tf.Session(config=tfconfig) as sess: model = cyclegan(sess, args) ### 記得要進去改 建立模型的地方 # model.train(args) if args.phase == 'train' else model.test(args) ### 有時間待修,有CYCLE的版本 model.train_kong(args) if args.phase == 'train' else model.test( args) ### 無Cycle,有D,D有concat
def main(_): if args.phase == 'train': args.checkpoint_dir = os.path.join(args.checkpoint_dir, args.exp_id) args.sample_dir = os.path.join(args.sample_dir, args.exp_id) args.log_dir = os.path.join(args.log_dir, args.exp_id) if not os.path.exists(args.log_dir): print(args.log_dir) os.mkdir(args.log_dir) if not os.path.exists(args.checkpoint_dir): os.makedirs(args.checkpoint_dir) if not os.path.exists(args.sample_dir): os.makedirs(args.sample_dir) with open(os.path.join(args.log_dir, 'training_config.txt'), 'w') as f: dic = vars(args) pp = pprint.PrettyPrinter(indent=1, width=80, depth=None, stream=f) pp.pprint(dic) else: if not os.path.exists(args.test_dir): os.makedirs(args.test_dir) with open(os.path.join(args.test_dir, 'config.txt'), 'w') as f: dic = vars(args) pp = pprint.PrettyPrinter(indent=1, width=80, depth=None, stream=f) pp.pprint(dic) tfconfig = tf.ConfigProto(allow_soft_placement=True) tfconfig.gpu_options.allow_growth = True with tf.Session(config=tfconfig) as sess: model = cyclegan(sess, args) model.train(args) if args.phase == 'train' \ else model.test(args)
def load_cg(): global cyclegan_model tfconfig = tf.ConfigProto(allow_soft_placement=True) tfconfig.gpu_options.allow_growth = True sess = tf.Session(config=tfconfig) cyclegan_model = cyclegan(sess, args) cyclegan_model.init_load(args)
def main(_): if not os.path.exists(args.checkpoint_dir): os.makedirs(args.checkpoint_dir) with tf.Session() as sess: model = cyclegan(sess, args) if args.phase == 'train': model.train(args) else: model.test(args)
def main(_): if not os.path.exists(args.checkpoint_dir): os.makedirs(args.checkpoint_dir) if not os.path.exists(args.sample_dir): os.makedirs(args.sample_dir) if not os.path.exists(args.test_dir): os.makedirs(args.test_dir) tfconfig = tf.ConfigProto(allow_soft_placement=True) tfconfig.gpu_options.allow_growth = True with tf.Session(config=tfconfig) as sess: model = cyclegan(sess, args) model.train(args) if args.phase == 'train' else model.test(args)
def main(_): if not os.path.exists(args.checkpoint_dir): os.makedirs(args.checkpoint_dir) if not os.path.exists(args.sample_dir): os.makedirs(args.sample_dir) if not os.path.exists(args.test_dir): os.makedirs(args.test_dir) tfconfig = tf.ConfigProto(allow_soft_placement=True) tfconfig.gpu_options.allow_growth = True with tf.Session(config=tfconfig) as sess: model = cyclegan(sess, args) model.train(args) if args.phase == 'train' \ else model.test(args)
def modification(): r = request # convert string of image data to uint8s #nparr = np.fromstring(r.data, np.uint8) # decode image #f = r.files['file'] file = r.files['upload'] img = Image.open(file) img.save('./datasets/makeup2/testA/input.jpg') #f = os.path.join(app.config['UPLOAD_FOLDER'], file.filename) #file.save(f) #f.save(secure_filename(f.filename)) #img = cv2.imdecode(nparr,cv2.IMREAD_COLOR) ''' tfconfig = tf.ConfigProto(allow_soft_placement=True) tfconfig.gpu_options.allow_growth = True with tf.Session(config=tfconfig) as sess: model = cyclegan(sess, args) if model.load(args.checkpoint_dir): print(" [*] Load SUCCESS") else: print(" [!] Load failed...") init_op = tf.global_variables_initializer() model.sess.run(init_op) out_var, in_var = (model.testB, model.test_A) img = np.expand_dims(img, axis=0) img = np.array(img).astype(np.float32) fake_img = model.sess.run(out_var, feed_dict={in_var: img}) save_images(fake_img, [1, 1], "./output.jpg") ## merge images ''' tfconfig = tf.ConfigProto(allow_soft_placement=True) tfconfig.gpu_options.allow_growth = True with tf.Session(config=tfconfig) as sess: model = cyclegan(sess, args) model.test(args) response = {'message': 'image received. size={}x{}'.format(1, 1)} print(response) # encode response using jsonpickle response_pickled = jsonpickle.encode(response) return Response(response=response_pickled, status=200, mimetype="application/json")
def main(_): if not os.path.exists(args.checkpoint_dir): os.makedirs(args.checkpoint_dir) if not os.path.exists(args.sample_dir): os.makedirs(args.sample_dir) if not os.path.exists(args.test_dir): os.makedirs(args.test_dir) with tf.Session() as sess: if args.debug == True: sess = tf_debug.LocalCLIDebugWrapperSession(sess) sess.add_tensor_filter("has_inf_or_nan", tf_debug.has_inf_or_nan) model = cyclegan(sess, args) # model.trainVAE(args) # for pretraining VAE model.train(args) # training full network
def main(_): if not os.path.exists(args.checkpoint_dir): os.makedirs(args.checkpoint_dir) if not os.path.exists(args.sample_dir): os.makedirs(args.sample_dir) if not os.path.exists(args.test_dir): os.makedirs(args.test_dir) tfconfig = tf.ConfigProto(allow_soft_placement=True) tfconfig.gpu_options.allow_growth = True # tfconfig.gpu_options.per_process_gpu_memory_fraction = 0.8 #saver = tf.train.Saver() with tf.Session(config=tfconfig) as sess: model = cyclegan(sess, args) model.train(args) if args.phase == 'train' \ else model.test(args)
def main(_): if not os.path.exists(args.checkpoint_dir): os.makedirs(args.checkpoint_dir) if not os.path.exists(args.sample_dir): os.makedirs(args.sample_dir) if not os.path.exists(args.test_dir): os.makedirs(args.test_dir) tfconfig = tf.ConfigProto(allow_soft_placement=True) tfconfig.gpu_options.allow_growth = True with tf.Session(config=tfconfig) as sess: model = cyclegan(sess, args) if args.phase == 'train': print('Training start') model.train(args) elif args.phase == 'test': model.test(args) elif args.phase == 'visualize': model.visualize(args.sample_dir, 0, True, args)
def main(_): #if(not os.path.exists('/data1/ICE_DATA/{}'.format(args.dataset_dir))): # download it # download_dataset(data_set=args.dataset_dir) if not os.path.exists(args.checkpoint_dir): os.makedirs(args.checkpoint_dir) if not os.path.exists(args.sample_dir): os.makedirs(args.sample_dir) if not os.path.exists(args.test_dir): os.makedirs(args.test_dir) tfconfig = tf.ConfigProto(allow_soft_placement=True) tfconfig.gpu_options.allow_growth = True with tf.Session(config=tfconfig) as sess: with tf.device('/device:GPU:1'): model = cyclegan(sess, args) model.train(args) if args.phase == 'train' \ else model.test(args)
def main(_): if not os.path.exists(args.checkpoint_dir): os.makedirs(args.checkpoint_dir) if not os.path.exists(args.sample_dir): os.makedirs(args.sample_dir) smoothenes_seperated = str(args.smootheness).split('.') model_dir = "%s_%s_%s_%s_%s" % (args.dataset_dir, args.fine_size, args.h_hops, smoothenes_seperated[0], smoothenes_seperated[1]) if not os.path.exists(args.sample_dir+"/"+model_dir): os.makedirs(args.sample_dir+"/"+model_dir) if not os.path.exists(args.test_dir): os.makedirs(args.test_dir) with tf.device('/gpu:{}'.format(args.gpu)): tfconfig = tf.ConfigProto(allow_soft_placement=True) tfconfig.gpu_options.allow_growth = True with tf.Session(config=tfconfig) as sess: model = cyclegan(sess, args) model.train(args) if args.phase == 'train' \ else model.test(args)
def main(_): if not os.path.exists(args.checkpoint_dir): os.makedirs(args.checkpoint_dir) if not os.path.exists(args.test_dir): os.makedirs(args.test_dir) tfconfig = tf.ConfigProto(allow_soft_placement=True) tfconfig.gpu_options.allow_growth = True with tf.Session(config=tfconfig) as sess: model = cyclegan(sess, args) #model 불러옴 OPTIONS = namedtuple( 'OPTIONS', 'batch_size image_size \ gf_dim df_dim output_c_dim is_training') model.options = OPTIONS._make( (args.batch_size, args.fine_size, args.ngf, args.ndf, args.output_nc, args.phase == 'train')) model._build_model() model.saver = tf.train.Saver() #for 문 while (1): print('py ready') #동호오빠가 넘겨줄 부분 (0 : 파일이름, 1:확장자, 2:표정, 3: more 버튼이 눌렸는지) lines = input().split(',') args.new_file = lines[0] + '.' + lines[1] args.which_expression = lines[2] args.more_button = lines[3] args.new_file_name = lines[0] + '_result.' + lines[1] if lines[0] == 99: break if model.load(): print(" [*] Load SUCCESS") else: print(" [*] Load Failed...") model.test(args)
### analysis gray_img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) for i in range(1): gray_img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) if gray_img.mean() < 130: img = adjust_gamma(img, 1.5) else: break return img if video == 'v': tfconfig = tf.ConfigProto(allow_soft_placement=True) tfconfig.gpu_options.allow_growth = True with tf.Session(config=tfconfig) as sess: model = cyclegan(sess, args2) model.loadModel() vc = cv2.VideoCapture('./data/video3.mp4') length = int(vc.get(cv2.CAP_PROP_FRAME_COUNT)) print('length :', length) if args["with_draw"] == 'True': cv2.namedWindow('show', 0) dir_name = str(now.year) + "_" + str(now.month) + "_" + str( now.day) + "_" + str(now.minute) + "_" + str( now.second) #+ "_" + str(idx) os.makedirs(dir_name) print(dir_name) for idx in range(length): img_bgr = vc.read()[1] if img_bgr is None: break
else: x_train = x_train.reshape(x_train.shape[0], img_rows, img_cols, 1) x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols, 1) input_shape = (img_rows, img_cols, 1) score = model.evaluate(x_train, y_train.reshape(-1, ), verbose=0) print('svhn Train loss:', score[0]) print('svhn Train accuracy:', score[1]) score = model.evaluate(x_test, y_test, verbose=0) print('svhn Test loss:', score[0]) print('svhn Test accuracy:', score[1]) print("begining cycle gan.....") with tf.Session(config=tfconfig) as sess: domain_adapation_model = cyclegan(sess, args) x_test_c = np.zeros(x_test.shape) batch_size = 2 batch_num = int(x_test.shape[0] / batch_size) for idx in range(batch_num): print(idx) x_test_c[idx * batch_size:(idx + 1) * batch_size] = domain_adapation_model.pix2pix_cylce_gan( args, x_test[idx * batch_size:(idx + 1) * batch_size]) score = model.evaluate(x_test_c, y_test, verbose=0) print('svhn Test loss after cycle gan:', score[0]) print('svhn Test accuracy after cycle gan:', score[1]) #x_train_c = domain_adapation_model.pix2pix_cylce_gan(args, x_train) #score = model.evaluate(x_train_c, y_train, verbose=0) #print('svhn Train loss after cycle gan:', score[0]) #print('svhn Train accuracy after cycle gan:', score[1])