File size: 2,935 Bytes
a89d9fd |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 |
# copyright (c) 2022 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.
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import math
import paddle
from paddle import ParamAttr
import paddle.nn as nn
import paddle.nn.functional as F
from ppocr.modeling.necks.rnn import Im2Seq, EncoderWithRNN, EncoderWithFC, SequenceEncoder, EncoderWithSVTR
from .rec_ctc_head import CTCHead
from .rec_sar_head import SARHead
class MultiHead(nn.Layer):
def __init__(self, in_channels, out_channels_list, **kwargs):
super().__init__()
self.head_list = kwargs.pop('head_list')
self.gtc_head = 'sar'
assert len(self.head_list) >= 2
for idx, head_name in enumerate(self.head_list):
name = list(head_name)[0]
if name == 'SARHead':
# sar head
sar_args = self.head_list[idx][name]
self.sar_head = eval(name)(in_channels=in_channels, \
out_channels=out_channels_list['SARLabelDecode'], **sar_args)
elif name == 'CTCHead':
# ctc neck
self.encoder_reshape = Im2Seq(in_channels)
neck_args = self.head_list[idx][name]['Neck']
encoder_type = neck_args.pop('name')
self.encoder = encoder_type
self.ctc_encoder = SequenceEncoder(in_channels=in_channels, \
encoder_type=encoder_type, **neck_args)
# ctc head
head_args = self.head_list[idx][name]['Head']
self.ctc_head = eval(name)(in_channels=self.ctc_encoder.out_channels, \
out_channels=out_channels_list['CTCLabelDecode'], **head_args)
else:
raise NotImplementedError(
'{} is not supported in MultiHead yet'.format(name))
def forward(self, x, targets=None):
ctc_encoder = self.ctc_encoder(x)
ctc_out = self.ctc_head(ctc_encoder, targets)
head_out = dict()
head_out['ctc'] = ctc_out
head_out['ctc_neck'] = ctc_encoder
# eval mode
if not self.training:
return ctc_out
if self.gtc_head == 'sar':
sar_out = self.sar_head(x, targets[1:])
head_out['sar'] = sar_out
return head_out
else:
return head_out
|