def inference(): graph = tf.Graph() with graph.as_default(): with tf.gfile.FastGFile('input_image.jpg', 'rb') as f: image_data = f.read() print("1") input_image = tf.image.decode_jpeg(image_data, channels=3) input_image = tf.image.resize_images(input_image, size=(FLAGS.image_size, FLAGS.image_size)) input_image = utils.convert2float(input_image) input_image.set_shape([FLAGS.image_size, FLAGS.image_size, 3]) print("2") with tf.gfile.FastGFile('./pretrained/apple2orange.pb', 'rb') as model_file: graph_def = tf.GraphDef() print("3") graph_def.ParseFromString(model_file.read()) [output_image ] = tf.import_graph_def(graph_def, input_map={'input_image': input_image}, return_elements=['output_image:0'], name='output') with tf.Session(graph=graph) as sess: generated = output_image.eval() with open(FLAGS.output, 'wb') as f: f.write(generated)
def translate(): graph = tf.Graph() with graph.as_default(): with tf.gfile.FastGFile(FLAGS.input_img, 'rb') as f: img = f.read() input_img = tf.image.decode_jpeg(img, channels=3) input_img = tf.image.resize_images(input_img, size=(FLAGS.img_size, FLAGS.img_size)) input_img = utils.convert2float(input_img) input_img.set_shape([FLAGS.img_size, FLAGS.img_size, 3]) with tf.gfile.FastGFile(FLAGS.model, 'rb') as model: graph_def = tf.GraphDef() graph_def.ParseFromString(model.read()) [output_img ] = tf.import_graph_def(graph_def, input_map={'input_image': input_img}, return_elements=['output_image:0'], name='output') with tf.Session(graph=graph) as sess: generated = output_img.eval() with open(FLAGS.output_img, 'wb') as f: f.write(generated)
def sample(): graph = tf.Graph() with graph.as_default(): cycle_gan = CycleGAN() with tf.gfile.FastGFile(IMG_PATH, 'r') as f: image_data = f.read() in_image = tf.image.decode_jpeg(image_data, channels=3) in_image = tf.image.resize_images(in_image, size=(128, 128)) in_image = utils.convert2float(in_image) in_image.set_shape([128, 128, 3]) cycle_gan = CycleGAN() cycle_gan.model() out_image = cycle_gan.sample(tf.expand_dims(in_image, 0)) with tf.Session(graph=graph) as sess: sess.run(tf.global_variables_initializer()) cycle_gan.saver.restore(sess, CKPT_PATH) generated = out_image.eval() samples_dir = 'samples' os.makedirs(samples_dir, exist_ok=True) samples_file = os.path.join(samples_dir, 'sample.jpg') with open(samples_file, 'wb') as f: f.write(generated)
def test(file): dataset = FLAGS.input.split("/")[1] + '/' test_name = FLAGS.input.split("/")[2] + '/' graph = tf.Graph() with graph.as_default(): print('Reading in image: ' + file) with tf.gfile.FastGFile(FLAGS.input + file, 'rb') as f: image_data = f.read() input_image = tf.image.decode_jpeg(image_data, channels=3) input_image = tf.image.resize_images(input_image, size=(FLAGS.image_size, FLAGS.image_size)) input_image = utils.convert2float(input_image) input_image.set_shape([FLAGS.image_size, FLAGS.image_size, 3]) with tf.gfile.FastGFile(FLAGS.model, 'rb') as model_file: graph_def = tf.GraphDef() graph_def.ParseFromString(model_file.read()) [output] = tf.import_graph_def(graph_def, input_map={'input_image': input_image}, return_elements=['output_image:0'], name='output') with tf.Session(graph=graph) as sess: generated = output.eval() with open(FLAGS.output + dataset + test_name + file, 'wb') as f: f.write(generated)
def inference(url="", outputpath="output.jpg", isurl=True, modelpath="zebra2horse.pb"): graph = tf.Graph() with graph.as_default(): if isurl: image_data = requests.get(url=url).content else: #print(url) with open(url, "rb") as f: image_data = f.read() input_image = tf.image.decode_jpeg(image_data, channels=3) input_image = tf.image.resize_images(input_image, size=(FLAGS.image_size, FLAGS.image_size)) input_image = utils.convert2float(input_image) input_image.set_shape([FLAGS.image_size, FLAGS.image_size, 3]) with tf.gfile.FastGFile(modelpath, 'rb') as model_file: graph_def = tf.GraphDef() graph_def.ParseFromString(model_file.read()) [output_image ] = tf.import_graph_def(graph_def, input_map={'input_image': input_image}, return_elements=['output_image:0'], name='output') with tf.Session(graph=graph) as sess: generated = output_image.eval() with open(outputpath, 'wb') as f: f.write(generated)
def inference(): graph = tf.Graph() #创建计算图 with graph.as_default(): with tf.gfile.FastGFile(FLAGS.input, 'rb') as f: image_data = f.read() input_image = tf.image.decode_jpeg(image_data, channels=3) #解码 input_image = tf.image.resize_images(input_image, size=(FLAGS.image_size, FLAGS.image_size)) input_image = utils.convert2float(input_image) input_image.set_shape([FLAGS.image_size, FLAGS.image_size, 3]) with tf.gfile.FastGFile(FLAGS.model, 'rb') as model_file: #导入模型 graph_def = tf.GraphDef() graph_def.ParseFromString(model_file.read()) [output_image ] = tf.import_graph_def(graph_def, input_map={'input_image': input_image}, return_elements=['output_image:0'], name='output') with tf.Session(graph=graph) as sess: generated = output_imag.eval() #计算output_image with open(FLAGS.output, 'wb') as f: #将计算结果写入output文件 f.write(generated)
def inference(): graph = tf.Graph() for ind in range(0,18): with graph.as_default(): input_image = r'E:\data\after_spm\031419464625\%d.jpg'%ind with tf.gfile.FastGFile(input_image, 'rb') as f: image_data = f.read() input_image = tf.image.decode_jpeg(image_data, channels=3) input_image = tf.image.resize_images(input_image, size=(FLAGS.image_size, FLAGS.image_size)) input_image = utils.convert2float(input_image) input_image.set_shape([FLAGS.image_size, FLAGS.image_size, 3]) with tf.gfile.FastGFile(FLAGS.model, 'rb') as model_file: graph_def = tf.GraphDef() graph_def.ParseFromString(model_file.read()) [output_image] = tf.import_graph_def(graph_def, input_map={'input_image': input_image}, return_elements=['output_image:0'], name='output') with tf.Session(graph=graph) as sess: generated = output_image.eval() output_image = 'E:\\data\\after_spm\\OUTPUT_REAL\\output_histogram_%d.jpg'%ind with open(output_image, 'wb') as f: f.write(generated)
def inference(model, name, artist, img_in, img_out, size=256): graph = tf.Graph() with graph.as_default(): with tf.gfile.FastGFile(img_in, 'rb') as f: image_data = f.read() input_image = tf.image.decode_jpeg(image_data, channels=3) input_image = tf.image.resize_images(input_image, size=(size, size)) input_image = utils.convert2float(input_image) input_image.set_shape([size, size, 3]) with tf.gfile.FastGFile(model, 'rb') as model_file: graph_def = tf.GraphDef() graph_def.ParseFromString(model_file.read()) [output_image] = tf.import_graph_def(graph_def, input_map={'input_image': input_image}, return_elements=['output_image:0'], name='output') with tf.Session(graph=graph) as sess: generated = output_image.eval() out_art = Image.open(io.BytesIO(generated)) draw_text(out_art, name.replace("$", " ")) draw_text(out_art, artist.replace("$", " "), bottom=False) out_art.save(img_out, "JPEG")
def inference(): graph = tf.Graph() with graph.as_default(): with tf.gfile.FastGFile(FLAGS.input, 'rb') as f: image_data = f.read() input_image = tf.image.decode_jpeg(image_data, channels=3) input_image = tf.image.transpose_image(input_image) input_image = tf.image.resize_images(input_image, size=(FLAGS.image_length, FLAGS.image_height)) # input_image = tf.image.resize_images(input_image, size=(FLAGS.image_height, FLAGS.image_length)) input_image = utils.convert2float(input_image) input_image.set_shape([FLAGS.image_length, FLAGS.image_height, 3]) # input_image.set_shape([FLAGS.image_height, FLAGS.image_length, 3]) with tf.gfile.FastGFile(FLAGS.model, 'rb') as model_file: graph_def = tf.GraphDef() graph_def.ParseFromString(model_file.read()) [output_image] = tf.import_graph_def(graph_def, input_map={'input_image': input_image}, return_elements=['output_image:0'], name='output') with tf.Session(graph=graph) as sess: generated = output_image.eval() with open(FLAGS.output, 'wb') as f: f.write(generated)
def _preprocess(self, image): image = tf.image.resize_images(image, size=(self.image_size[0], self.image_size[1])) image = utils.convert2float(image) image.set_shape([self.image_size[0], self.image_size[1], 3]) return image
def color(src, dst): MODEL = 'pretrained/sketch2render.pb' IMG_SIZE = 256 graph = tf.Graph() with graph.as_default(): with tf.io.gfile.GFile(src, 'rb') as f: image_data = f.read() input_image = tf.image.decode_jpeg(image_data, channels=3) input_image = tf.image.resize(input_image, size=(IMG_SIZE, IMG_SIZE)) input_image = utils.convert2float(input_image) input_image.set_shape([IMG_SIZE, IMG_SIZE, 3]) with tf.io.gfile.GFile(MODEL, 'rb') as model_file: # graph_def = tf.Graph() graph_def = tf.compat.v1.GraphDef() graph_def.ParseFromString(model_file.read()) [output_image] = tf.import_graph_def(graph_def, input_map={'input_image': input_image}, return_elements=['output_image:0'], name='output') with tf.compat.v1.Session(graph=graph) as sess: generated = output_image.eval() with open(dst, 'wb') as f: f.write(generated)
def sample(): """Translate image to image (currently only support image with size 128x128)""" graph = tf.Graph() with graph.as_default(): with tf.gfile.FastGFile(FLAGS.input, 'r') as f: image_data = f.read() input_image = tf.image.decode_jpeg(image_data, channels=3) input_image = tf.image.resize_images(input_image, size=(128, 128)) input_image = utils.convert2float(input_image) input_image.set_shape([128, 128, 3]) with tf.gfile.FastGFile(FLAGS.model, 'rb') as model_file: graph_def = tf.GraphDef() graph_def.ParseFromString(model_file.read()) [output_image ] = tf.import_graph_def(graph_def, input_map={'input_image': input_image}, return_elements=['output_image:0'], name='apple2orange') with tf.Session(graph=graph) as sess: generated = output_image.eval() with open(FLAGS.output, 'wb') as f: f.write(generated)
def main(): picF = "picF" files = os.listdir(picF)[:1] sess = tf.InteractiveSession() global_step = tf.Variable(2501, name="global_step", trainable=False) sess.run(tf.global_variables_initializer()) img = tf.read_file(os.path.join(picF, files[0])) img = tf.image.decode_jpeg(img) #img = utils.convert2float(img) img = tf.expand_dims(img, axis=0) tf.summary.image('real', img) tf.summary.scalar('test', global_step) outimg = utils.convert2float(img) tf.summary.image('out', utils.batch_convert2int(outimg)) sdf = outimg.eval() print(sdf) sdf = utils.batch_convert2int(outimg).eval() print(sdf) summary_op = tf.summary.merge_all() train_writer = tf.summary.FileWriter('C:\\log') summary, _, s, outimg2 = sess.run([summary_op, img, global_step, outimg]) train_writer.add_summary(summary) train_writer.flush() sess.close()
def inference(): test = os.listdir(FLAGS.test_path) graph = tf.Graph() for index in range(len(test)): with graph.as_default(): with tf.gfile.FastGFile(FLAGS.test_path + "/" + test[index], 'rb') as f: image_data = f.read() input_image = tf.image.decode_jpeg(image_data, channels=3) input_image = tf.image.resize_images(input_image, size=(FLAGS.image_size, FLAGS.image_size)) input_image = utils.convert2float(input_image) input_image.set_shape([FLAGS.image_size, FLAGS.image_size, 3]) with tf.gfile.FastGFile(FLAGS.model, 'rb') as model_file: graph_def = tf.GraphDef() graph_def.ParseFromString(model_file.read()) [output_image ] = tf.import_graph_def(graph_def, input_map={'input_image': input_image}, return_elements=['output_image:0'], name='output') with tf.Session(graph=graph) as sess: generated = output_image.eval() with open(FLAGS.output + test[index], 'wb') as f: f.write(generated)
def inference(): graph = tf.Graph() with graph.as_default(): with tf.gfile.FastGFile(FLAGS.input, 'rb') as f: image_data = f.read() input_image = tf.image.decode_jpeg(image_data, channels=3) input_image = tf.image.resize_images(input_image, size=(FLAGS.image_size, FLAGS.image_size)) input_image = utils.convert2float(input_image) input_image.set_shape([FLAGS.image_size, FLAGS.image_size, 3]) with tf.gfile.FastGFile(FLAGS.model, 'rb') as model_file: graph_def = tf.GraphDef() graph_def.ParseFromString(model_file.read()) for node in graph_def.node: if node.op == 'RefSwitch': node.op = 'Switch' for index in range(len(node.input)): if 'moving_' in node.input[index]: node.input[index] = node.input[index] + '/read' elif node.op == 'AssignSub': node.op = 'Sub' if 'use_locking' in node.attr: del node.attr['use_locking'] [output_image] = tf.import_graph_def(graph_def, input_map={'input_image': input_image}, return_elements=['output_image:0'], name='output') with tf.Session(graph=graph) as sess: generated = output_image.eval() with open(FLAGS.output, 'wb') as f: f.write(generated)
def _preprocess(self, image): image = tf.image.resize_images(image, size=(self.image_length, self.image_height)) image = utils.convert2float(image) image.set_shape([self.image_length, self.image_height, 3]) return image
def inference(): graph = tf.Graph() with tf.Session(graph=graph) as sess: with graph.as_default(): for input in imgs: output = input[0:-4] + '_f.jpg' with tf.gfile.FastGFile(input, 'rb') as f: image_data = f.read() input_image = tf.image.decode_jpeg(image_data, channels=3) input_image = tf.image.resize_images(input_image, size=(image_size, image_size)) input_image = utils.convert2float(input_image) input_image.set_shape([image_size, image_size, 3]) with tf.gfile.FastGFile(model, 'rb') as model_file: graph_def = tf.GraphDef() graph_def.ParseFromString(model_file.read()) [output_image] = tf.import_graph_def(graph_def, input_map={'input_image': input_image}, return_elements=['output_image:0'], name='output') generated = output_image.eval() with open(output, 'wb') as f: f.write(generated)
def inference(): graph = tf.Graph() with graph.as_default(): with tf.gfile.FastGFile(FLAGS.model, 'rb') as model_file: graph_def = tf.GraphDef() graph_def.ParseFromString(model_file.read()) outputs = [] files = os.listdir(FLAGS.input) for filename in files: with tf.gfile.FastGFile(FLAGS.input + '/' + filename, 'rb') as f: image_data = f.read() if FLAGS.direction == 'XtoY': input_image = tf.image.decode_jpeg(image_data, channels=1) input_image = tf.image.resize_images( input_image, size=(FLAGS.image_size, FLAGS.image_size)) input_image = utils.convert2float(input_image) input_image.set_shape( [FLAGS.image_size, FLAGS.image_size, 1]) elif FLAGS.direction == 'YtoX': input_image = tf.image.decode_jpeg(image_data, channels=3) input_image = tf.image.resize_images( input_image, size=(FLAGS.image_size, FLAGS.image_size)) input_image = utils.convert2float(input_image) input_image.set_shape( [FLAGS.image_size, FLAGS.image_size, 3]) if FLAGS.direction == 'XtoY': [output_image] = tf.import_graph_def( graph_def, input_map={'input_image_X': input_image}, return_elements=['output_image:0'], name='output') elif FLAGS.direction == 'YtoX': [output_image] = tf.import_graph_def( graph_def, input_map={'input_image_Y': input_image}, return_elements=['output_image:0'], name='output') outputs = outputs + [output_image] with tf.Session(graph=graph) as sess: for (output_image, filename) in zip(outputs, files): generated = output_image.eval() with open(FLAGS.output + '/' + filename, 'wb') as f: f.write(generated)
def _preprocess(self, image): image = tf.image.resize_images(image, size=(self.image_size, self.image_size)) image = utils.convert2float(image) if self.name == 'X': image.set_shape([self.image_size, self.image_size, 1]) elif self.name == 'Y': image.set_shape([self.image_size, self.image_size, 3]) return image
def _preprocess(self, image): #image = tf.image.resize_images(image, size=(self.image_size, self.image_size)) image = tf.image.resize_images(image, size=(self.image_height, self.image_width)) image = utils.convert2float(image) #image.set_shape([self.image_size, self.image_size, 3]) image.set_shape([self.image_height, self.image_width, 3]) return image
def inference(): graph = tf.Graph() with graph.as_default(): with tf.gfile.FastGFile(FLAGS.input1, 'rb') as f: image_data = f.read() input_image1 = tf.image.decode_jpeg(image_data, channels=3) input_image1 = tf.image.resize_images(input_image1, size=(FLAGS.image_size, FLAGS.image_size)) input_image1 = utils.convert2float(input_image1) input_image1.set_shape([FLAGS.image_size, FLAGS.image_size, 3]) with tf.gfile.FastGFile(FLAGS.input2, 'rb') as f: image_data = f.read() input_image2 = tf.image.decode_jpeg(image_data, channels=3) input_image2 = tf.image.resize_images(input_image2, size=(FLAGS.image_size, FLAGS.image_size)) input_image2 = utils.convert2float(input_image2) input_image2.set_shape([FLAGS.image_size, FLAGS.image_size, 3]) with tf.gfile.FastGFile(FLAGS.model, 'rb') as model_file: graph_def = tf.GraphDef() graph_def.ParseFromString(model_file.read()) [output_image1, output_image2] = tf.import_graph_def( graph_def, input_map={ 'input_image1': input_image1, 'input_image2': input_image2 }, return_elements=['output_image1:0', 'output_image2:0'], name='output') with tf.Session(graph=graph) as sess: generated1 = output_image1.eval() with open(FLAGS.output1, 'wb') as f: f.write(generated1) generated2 = output_image2.eval() with open(FLAGS.output2, 'wb') as f: f.write(generated2)
def _preprocess(self, image): image = tf.transpose(image, [1, 0, 2]) x = tf.random_uniform([], 0, 1120, dtype = tf.int32) y = tf.random_uniform([], 0, 640, dtype = tf.int32) image = tf.image.crop_to_bounding_box(image, x, y, self.image_size, self.image_size) image = tf.contrib.image.rotate(image, -math.pi/2) #image = tf.image.convert_image_dtype(image, dtype=tf.float32) image = utils.convert2float(tf.cast(image, tf.float32)) image.set_shape([self.image_size, self.image_size, 3]) return image
def _preprocess(self, image): """ 读取并对TFrecords文件解码 若需要处理非L/RGB/RGBA类型的图像,请自行添加代码 python Image 读入的图像按照[height,weight,depth]维度排列 Return: image: 3D tensor [image_width, image_height, image_depth] """ if self.image_mode == 'L': image = tf.reshape(image, [self.image_width, self.image_height, 1]) image = utils.convert2float(image) elif self.image_mode == 'RGB': image = tf.reshape(image, [self.image_width, self.image_height, 3]) image = utils.convert2float(image) elif self.image_mode == 'RGBA': image = tf.reshape(image, [self.image_width, self.image_height, 4]) image = utils.convert2float(image) #image = tf.cast(image,tf.float32)*(1./255)-0.5 else: print('The image mode must be L/RGB/RGBA!') sys.exit() return image
def main(): sess = tf.InteractiveSession() testimg = np.asarray([[0,127,255], [20,128,127], [255,100,100]],dtype=np.float) print(testimg) img = tf.convert_to_tensor(testimg,dtype=tf.uint8) print(img) img = tf.expand_dims(img,axis=2) print(img) print(img.eval()) #img = tf.image.convert_image_dtype(img,dtype= tf.float32) img = utils.convert2float(img) #img = (img/127.5)-1.0 print(img) print(img.eval()) img = (img +1.0)/2.0 img = tf.image.convert_image_dtype(img,dtype=tf.uint8) print(img) print(img.eval())
def inference(): graph = tf.Graph() # visualization staff yw3025 index = open(FLAGS.input + "/" + FLAGS.direction + "index.html", "w") index.write("<html><body><table><tr>") index.write("<th>name</th><th>input</th><th>output</th></tr>") # batch inference staff yw3025 for file in os.listdir(FLAGS.input): filename = FLAGS.input + "/" + file print(filename) with graph.as_default(): with tf.gfile.FastGFile(filename, 'rb') as f: image_data = f.read() input_image = tf.image.decode_jpeg(image_data, channels=3) input_image = tf.image.resize_images(input_image, size=(FLAGS.image_size, FLAGS.image_size)) input_image = utils.convert2float(input_image) input_image.set_shape([FLAGS.image_size, FLAGS.image_size, 3]) with tf.gfile.FastGFile(FLAGS.model, 'rb') as model_file: graph_def = tf.GraphDef() graph_def.ParseFromString(model_file.read()) [output_image ] = tf.import_graph_def(graph_def, input_map={'input_image': input_image}, return_elements=['output_image:0'], name='output') with tf.Session(graph=graph) as sess: generated = output_image.eval() with open(FLAGS.output + "/" + FLAGS.direction + file, 'wb') as f: f.write(generated) # visualization staff yw3025 index.write("<td>%s</td>" % filename) index.write("<td><img src='%s'></td>" % file) index.write("<td><img src='%s'></td>" % (FLAGS.direction + file)) index.write("</tr>") print("processing" + filename)
def main(): modelfile = 'model\\model.ckpt-2000' picF = "picF" files = os.listdir(picF) sess = tf.InteractiveSession() ge = Generator('G', is_training=False) x = tf.placeholder(dtype=tf.float32, shape=(5, 270, 480, 3)) out = ge(x) sess.run(tf.global_variables_initializer()) saver = tf.train.Saver(var_list=ge.variables) v = ge.variables[0] print(v.name) saver.restore(sess, modelfile) print(v.eval()) for pic in files: img = tf.read_file(os.path.join(picF, pic)) img = tf.image.decode_jpeg(img) img = utils.convert2float(img) img = tf.expand_dims(img, axis=0) shape = tf.shape(img).eval() img.set_shape(shape) out = ge(img) out = tf.unstack(out)[0] out = utils.convert2int(out) out = tf.image.encode_jpeg(out) out = out.eval() with tf.gfile.GFile(os.path.join('out\\picF', pic), 'wb') as fw: fw.write(out) fw.flush() sess.close()
def inference(): graph = tf.Graph() with graph.as_default(): with tf.gfile.FastGFile(FLAGS.model, 'rb') as model_file: graph_def = tf.GraphDef() graph_def.ParseFromString(model_file.read()) file_list = data_reader(FLAGS.input) whole = len(file_list) cnt = 0 with tf.Session(graph=graph) as sess: for file in file_list: with tf.gfile.FastGFile(file, 'rb') as f: image_data = f.read() input_image = tf.image.decode_jpeg(image_data, channels=3) input_image = tf.image.resize_images( input_image, size=(FLAGS.image_size, FLAGS.image_size)) input_image = utils.convert2float(input_image) input_image.set_shape( [FLAGS.image_size, FLAGS.image_size, 3]) #input_image_list.append(input_image) print cnt [output_image] = tf.import_graph_def( graph_def, input_map={'input_image': input_image}, return_elements=['output_image:0'], name='output') print cnt generated = output_image.eval() print cnt output_file_name = file.split('/')[-1] with open(FLAGS.output + '/fake_{}'.format(output_file_name), 'wb') as f: f.write(generated) cnt += 1 if cnt / whole > 0.05: print cnt / whole, 'done'
def _preprocess(self, image): image = tf.image.resize_images(image, size=(self.image_size, self.image_size)) image = convert2float(image) image.set_shape([self.image_size, self.image_size, 3]) return image
def inference(): path_src_test_pic = FLAGS.Test_input path_dst_test_dir = FLAGS.Test_output temp_img_list = os.listdir(path_src_test_pic) image_size = FLAGS.image_size path_model_dir = FLAGS.Model_dir model_name_list = os.listdir(path_model_dir) Total_count_model = len(model_name_list) Total_count_img = len(temp_img_list) count_model = 1 ################################################################################# for model_name in model_name_list: path_model_now_use = os.path.join(path_model_dir, model_name) print("Model Progress :" + str(count_model) + "/" + str(Total_count_model)) print("Now use Model is " + model_name) count_model += 1 count_img = 1 model_name = model_name.split(".") model_name = model_name[0] output_by_model_dir = os.path.join(path_dst_test_dir, model_name) if not os.path.isdir(output_by_model_dir): os.mkdir(output_by_model_dir) print("Image will save in the path of directory : " + output_by_model_dir) for img in temp_img_list: graph = tf.Graph() start = time() with graph.as_default(): temp_input = os.path.join(path_src_test_pic, img) temp_output = os.path.join(output_by_model_dir, img) with tf.gfile.FastGFile(temp_input, 'rb') as f: image_data = f.read() input_image = tf.image.decode_jpeg(image_data, channels=3) input_image = tf.image.resize_images(input_image, size=(image_size, image_size)) input_image = utils.convert2float(input_image) input_image.set_shape([image_size, image_size, 3]) with tf.gfile.FastGFile(path_model_now_use, 'rb') as model_file: graph_def = tf.GraphDef() graph_def.ParseFromString(model_file.read()) [output_image] = tf.import_graph_def( graph_def, input_map={'input_image': input_image}, return_elements=['output_image:0'], name='output') with tf.Session(graph=graph) as sess: generated = output_image.eval() with open(temp_output, 'wb') as f: f.write(generated) print(str(count_img) + "/" + str(Total_count_img)) End = time() t = End - start print(str(t) + " sec") count_img += 1