exam-segment-django/segment/sheet_resolve/tools/tf_sess.py

# @Author  : lightXu
# @File    : tf_sess.py
# @Time    : 2018/11/23 0023 下午 14:26
import os
import tensorflow as tf
from segment.sheet_resolve.lib.model.config import cfg
from segment.sheet_resolve.lib.nets.resnet_v1 import resnetv1
from segment.sheet_resolve.tools.tf_settings import model_dict


class TfSess:

    def __init__(self, resolve_type):
        model_info = model_dict[resolve_type]
        self.graph = tf.Graph()
        with self.graph.as_default():
            cfg.TEST.HAS_RPN = True  # Use RPN for proposals
            # set config
            tfconfig = tf.ConfigProto(allow_soft_placement=True)
            tfconfig.gpu_options.allow_growth = True
            # tfconfig.gpu_options.per_process_gpu_memory_fraction = 0.4

            self.net = resnetv1(num_layers=101)
            self.net.create_architecture("TEST", len(model_info['classes']),
                                         tag='default', anchor_scales=model_info['anchor_scales'],
                                         anchor_ratios=model_info['anchor_ratios'])

            self.saver = tf.train.Saver()

        tfmodel = model_info['path']

        if not os.path.isfile(tfmodel + '.meta'):
            raise IOError('{:s} not found.'.format(tfmodel + '.meta'))

        self.sess = tf.Session(config=tfconfig, graph=self.graph)
        # self.sess.as_default()
        # try:
        #     self.saver.restore(self.sess, tfmodel)
        # except Exception:
        #     traceback.print_exc()

        with self.sess.as_default():
            with self.graph.as_default():
                self.saver.restore(self.sess, tfmodel)  # 从恢复点恢复参数

        print('\n    Loaded network {:s}\n'.format(tfmodel))

    def sess_close(self):
        self.sess.close()


class SsdSess:
    def __init__(self, ssd_type):
        model_info = model_dict[ssd_type]

        graph = tf.Graph()
        with graph.as_default():
            od_graph_def = tf.GraphDef()
            with tf.gfile.GFile(model_info['path'], 'rb') as fid:
                serialized_graph = fid.read()
                od_graph_def.ParseFromString(serialized_graph)
                tf.import_graph_def(od_graph_def, name='')

        self.sess = tf.Session(graph=graph)

        with graph.as_default():
            with self.sess.as_default():
                    self.graph = tf.get_default_graph()

        # with self.sess.as_default():
        #         self.graph = graph

        print('\n    Loaded network {:s}\n'.format(model_info['path']))

    def sess_close(self):
        self.sess.close()