train.py

#  Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserve.
#
#Licensed under the Apache License, Version 2.0 (the "License");
#you may not use this file except in compliance with the License.
#You may obtain a copy of the License at
#
#    http://www.apache.org/licenses/LICENSE-2.0
#
#Unless required by applicable law or agreed to in writing, software
#distributed under the License is distributed on an "AS IS" BASIS,
#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
#See the License for the specific language governing permissions and
#limitations under the License.
import sys

import paddle.v2 as paddle
import paddle.v2.fluid as fluid
import numpy as np


def random_reader(num_class):
    def reader():
        sequence_len = np.random.randint(5, 10)
        yield np.random.uniform(0.1, 1, [1, 512, 512]), np.random.randint(
            0, num_class + 1, [sequence_len])

    return reader


def ocr_conv(input, num, with_bn):
    assert (num % 4 == 0)

    def conv_block(input, filter_size, group_size, with_bn):
        return fluid.nets.img_conv_group(
            input=input,
            conv_num_filter=[filter_size] * group_size,
            pool_size=2,
            pool_stride=2,
            conv_padding=1,
            conv_filter_size=3,
            conv_act='relu',
            conv_with_batchnorm=with_bn,
            pool_type='max')

    conv1 = conv_block(input, 16, (num / 4), with_bn)
    conv2 = conv_block(conv1, 32, (num / 4), with_bn)
    conv3 = conv_block(conv2, 64, (num / 4), with_bn)
    conv4 = conv_block(conv3, 128, (num / 4), with_bn)
    return conv4


num_classes = 9054
data_shape = [1, 512, 512]

images = fluid.layers.data(name='pixel', shape=data_shape, dtype='float32')
label = fluid.layers.data(name='label', shape=[1], dtype='int64')

# encoder part
conv_features = ocr_conv(images, 8, True)

sliced_feature = fluid.layers.im2sequence(
    input=conv_features, stride=[1, 1], filter_size=[1, 3])

# TODO(wanghaoshuang): repaced by GRU
gru_forward, _ = fluid.layers.dynamic_lstm(input=sliced_feature, size=3 * 128)
gru_backward, _ = fluid.layers.dynamic_lstm(
    input=sliced_feature, size=3 * 128, is_reverse=True)

fc_out = fluid.layers.fc(input=[gru_forward, gru_backward],
                         size=num_classes + 1)

cost = fluid.layers.warpctc(
    input=fc_out,
    label=label,
    size=num_classes + 1,
    blank=num_classes,
    norm_by_times=True)
avg_cost = fluid.layers.mean(x=cost)

# TODO(wanghaoshuang): set clipping
optimizer = fluid.optimizer.Momentum(
    learning_rate=((1.0e-3) / 16), momentum=0.9)
opts = optimizer.minimize(cost)

decoded_out = fluid.layers.ctc_greedy_decoder(input=fc_out, blank=num_classes)
error_evaluator = fluid.evaluator.EditDistance(input=decoded_out, label=label)

BATCH_SIZE = 16
PASS_NUM = 1

# TODO(wanghaoshuang): replaced by correct data reader
train_reader = paddle.batch(
    paddle.reader.shuffle(
        random_reader(num_classes), buf_size=128 * 10),
    batch_size=BATCH_SIZE)

place = fluid.CPUPlace()
exe = fluid.Executor(place)
feeder = fluid.DataFeeder(place=place, feed_list=[images, label])
exe.run(fluid.default_startup_program())

for pass_id in range(PASS_NUM):
    error_evaluator.reset(exe)
    for data in train_reader():
        loss, error = exe.run(fluid.default_main_program(),
                              feed=feeder.feed(data),
                              fetch_list=[avg_cost] + error_evaluator.metrics)
        pass_error = error_evaluator.eval(exe)
        print "loss: %s;  distance error: %s; pass_dis_error: %s;" % (
            str(loss), str(error), str(pass_error))