magic-pdf 1.2.2__py3-none-any.whl → 1.3.0__py3-none-any.whl

magic_pdf/model/sub_modules/ocr/paddleocr2pytorch/pytorchocr/modeling/heads/__init__.py

@@ -0,0 +1,43 @@
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# 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.
+
+__all__ = ["build_head"]
+
+
+def build_head(config, **kwargs):
+    # det head
+    from .det_db_head import DBHead, PFHeadLocal
+
+    # rec head
+    from .rec_ctc_head import CTCHead
+    from .rec_multi_head import MultiHead
+
+    # cls head
+    from .cls_head import ClsHead
+
+    support_dict = [
+        "DBHead",
+        "CTCHead",
+        "ClsHead",
+        "MultiHead",
+        "PFHeadLocal",
+    ]
+
+    module_name = config.pop("name")
+    char_num = config.pop("char_num", 6625)
+    assert module_name in support_dict, Exception(
+        "head only support {}".format(support_dict)
+    )
+    module_class = eval(module_name)(**config, **kwargs)
+    return module_class

magic_pdf/model/sub_modules/ocr/paddleocr2pytorch/pytorchocr/modeling/heads/cls_head.py

@@ -0,0 +1,23 @@
+import torch
+import torch.nn.functional as F
+from torch import nn
+
+
+class ClsHead(nn.Module):
+    """
+    Class orientation
+    Args:
+        params(dict): super parameters for build Class network
+    """
+
+    def __init__(self, in_channels, class_dim, **kwargs):
+        super(ClsHead, self).__init__()
+        self.pool = nn.AdaptiveAvgPool2d(1)
+        self.fc = nn.Linear(in_channels, class_dim, bias=True)
+
+    def forward(self, x):
+        x = self.pool(x)
+        x = torch.reshape(x, shape=[x.shape[0], x.shape[1]])
+        x = self.fc(x)
+        x = F.softmax(x, dim=1)
+        return x

magic_pdf/model/sub_modules/ocr/paddleocr2pytorch/pytorchocr/modeling/heads/det_db_head.py

@@ -0,0 +1,109 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from ..common import Activation
+from ..backbones.det_mobilenet_v3 import ConvBNLayer
+
+class Head(nn.Module):
+    def __init__(self, in_channels, **kwargs):
+        super(Head, self).__init__()
+        self.conv1 = nn.Conv2d(
+            in_channels=in_channels,
+            out_channels=in_channels // 4,
+            kernel_size=3,
+            padding=1,
+            bias=False)
+        self.conv_bn1 = nn.BatchNorm2d(
+            in_channels // 4)
+        self.relu1 = Activation(act_type='relu')
+
+        self.conv2 = nn.ConvTranspose2d(
+            in_channels=in_channels // 4,
+            out_channels=in_channels // 4,
+            kernel_size=2,
+            stride=2)
+        self.conv_bn2 = nn.BatchNorm2d(
+            in_channels // 4)
+        self.relu2 = Activation(act_type='relu')
+
+        self.conv3 = nn.ConvTranspose2d(
+            in_channels=in_channels // 4,
+            out_channels=1,
+            kernel_size=2,
+            stride=2)
+
+    def forward(self, x, return_f=False):
+        x = self.conv1(x)
+        x = self.conv_bn1(x)
+        x = self.relu1(x)
+        x = self.conv2(x)
+        x = self.conv_bn2(x)
+        x = self.relu2(x)
+        if return_f is True:
+            f = x
+        x = self.conv3(x)
+        x = torch.sigmoid(x)
+        if return_f is True:
+            return x, f
+        return x
+
+
+class DBHead(nn.Module):
+    """
+    Differentiable Binarization (DB) for text detection:
+        see https://arxiv.org/abs/1911.08947
+    args:
+        params(dict): super parameters for build DB network
+    """
+
+    def __init__(self, in_channels, k=50, **kwargs):
+        super(DBHead, self).__init__()
+        self.k = k
+        binarize_name_list = [
+            'conv2d_56', 'batch_norm_47', 'conv2d_transpose_0', 'batch_norm_48',
+            'conv2d_transpose_1', 'binarize'
+        ]
+        thresh_name_list = [
+            'conv2d_57', 'batch_norm_49', 'conv2d_transpose_2', 'batch_norm_50',
+            'conv2d_transpose_3', 'thresh'
+        ]
+        self.binarize = Head(in_channels, **kwargs)# binarize_name_list)
+        self.thresh = Head(in_channels, **kwargs)#thresh_name_list)
+
+    def step_function(self, x, y):
+        return torch.reciprocal(1 + torch.exp(-self.k * (x - y)))
+
+    def forward(self, x):
+        shrink_maps = self.binarize(x)
+        return {'maps': shrink_maps}
+
+
+class LocalModule(nn.Module):
+    def __init__(self, in_c, mid_c, use_distance=True):
+        super(self.__class__, self).__init__()
+        self.last_3 = ConvBNLayer(in_c + 1, mid_c, 3, 1, 1, act='relu')
+        self.last_1 = nn.Conv2d(mid_c, 1, 1, 1, 0)
+
+    def forward(self, x, init_map, distance_map):
+        outf = torch.cat([init_map, x], dim=1)
+        # last Conv
+        out = self.last_1(self.last_3(outf))
+        return out
+
+class PFHeadLocal(DBHead):
+    def __init__(self, in_channels, k=50, mode='small', **kwargs):
+        super(PFHeadLocal, self).__init__(in_channels, k, **kwargs)
+        self.mode = mode
+
+        self.up_conv = nn.Upsample(scale_factor=2, mode="nearest")
+        if self.mode == 'large':
+            self.cbn_layer = LocalModule(in_channels // 4, in_channels // 4)
+        elif self.mode == 'small':
+            self.cbn_layer = LocalModule(in_channels // 4, in_channels // 8)
+
+    def forward(self, x, targets=None):
+        shrink_maps, f = self.binarize(x, return_f=True)
+        base_maps = shrink_maps
+        cbn_maps = self.cbn_layer(self.up_conv(f), shrink_maps, None)
+        cbn_maps = F.sigmoid(cbn_maps)
+        return {'maps': 0.5 * (base_maps + cbn_maps), 'cbn_maps': cbn_maps}

magic_pdf/model/sub_modules/ocr/paddleocr2pytorch/pytorchocr/modeling/heads/rec_ctc_head.py

@@ -0,0 +1,54 @@
+import torch.nn.functional as F
+from torch import nn
+
+
+class CTCHead(nn.Module):
+    def __init__(
+        self,
+        in_channels,
+        out_channels=6625,
+        fc_decay=0.0004,
+        mid_channels=None,
+        return_feats=False,
+        **kwargs
+    ):
+        super(CTCHead, self).__init__()
+        if mid_channels is None:
+            self.fc = nn.Linear(
+                in_channels,
+                out_channels,
+                bias=True,
+            )
+        else:
+            self.fc1 = nn.Linear(
+                in_channels,
+                mid_channels,
+                bias=True,
+            )
+            self.fc2 = nn.Linear(
+                mid_channels,
+                out_channels,
+                bias=True,
+            )
+
+        self.out_channels = out_channels
+        self.mid_channels = mid_channels
+        self.return_feats = return_feats
+
+    def forward(self, x, labels=None):
+        if self.mid_channels is None:
+            predicts = self.fc(x)
+        else:
+            x = self.fc1(x)
+            predicts = self.fc2(x)
+
+        if self.return_feats:
+            result = (x, predicts)
+        else:
+            result = predicts
+
+        if not self.training:
+            predicts = F.softmax(predicts, dim=2)
+            result = predicts
+
+        return result

magic_pdf/model/sub_modules/ocr/paddleocr2pytorch/pytorchocr/modeling/heads/rec_multi_head.py

@@ -0,0 +1,58 @@
+from torch import nn
+
+from ..necks.rnn import Im2Seq, SequenceEncoder
+from .rec_ctc_head import CTCHead
+
+
+class FCTranspose(nn.Module):
+    def __init__(self, in_channels, out_channels, only_transpose=False):
+        super().__init__()
+        self.only_transpose = only_transpose
+        if not self.only_transpose:
+            self.fc = nn.Linear(in_channels, out_channels, bias=False)
+
+    def forward(self, x):
+        if self.only_transpose:
+            return x.permute([0, 2, 1])
+        else:
+            return self.fc(x.permute([0, 2, 1]))
+
+
+class MultiHead(nn.Module):
+    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":
+                pass
+
+            elif name == "NRTRHead":
+                pass
+            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.ctc_encoder = SequenceEncoder(
+                    in_channels=in_channels, encoder_type=encoder_type, **neck_args
+                )
+                # ctc head
+                head_args = self.head_list[idx][name].get("Head", {})
+                if head_args is None:
+                    head_args = {}
+
+                self.ctc_head = CTCHead(
+                    in_channels=self.ctc_encoder.out_channels,
+                    out_channels=out_channels_list["CTCLabelDecode"],
+                    **head_args,
+                )
+            else:
+                raise NotImplementedError(f"{name} is not supported in MultiHead yet")
+
+    def forward(self, x, data=None):
+        ctc_encoder = self.ctc_encoder(x)
+        return self.ctc_head(ctc_encoder)

magic_pdf/model/sub_modules/ocr/paddleocr2pytorch/pytorchocr/modeling/necks/__init__.py

@@ -0,0 +1,29 @@
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# 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.
+
+__all__ = ["build_neck"]
+
+
+def build_neck(config):
+    from .db_fpn import DBFPN, LKPAN, RSEFPN
+    from .rnn import SequenceEncoder
+
+    support_dict = ["DBFPN", "SequenceEncoder", "RSEFPN", "LKPAN"]
+
+    module_name = config.pop("name")
+    assert module_name in support_dict, Exception(
+        "neck only support {}".format(support_dict)
+    )
+    module_class = eval(module_name)(**config)
+    return module_class