paddlex 3.0.0rc0__py3-none-any.whl → 3.0.1__py3-none-any.whl

paddlex/inference/models/open_vocabulary_segmentation/results/sam_result.py

@@ -1,4 +1,4 @@
-# copyright (c) 2024 PaddlePaddle Authors. All Rights Reserve.
+# Copyright (c) 2024 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.
@@ -12,23 +12,25 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import os
-import cv2
+import copy
+import random
+
 import numpy as np
-import copy, random
-import PIL
-from PIL import Image, ImageDraw, ImageFont
-from ....common.result import BaseCVResult, StrMixin, JsonMixin
+from PIL import Image
 
+from .....utils.deps import function_requires_deps, is_dep_available
+from ....common.result import BaseCVResult, JsonMixin
 from ....utils.color_map import get_colormap
-from ....common.result import BaseCVResult
+
+if is_dep_available("opencv-contrib-python"):
+    import cv2
 
 
+@function_requires_deps("opencv-contrib-python")
 def draw_segm(im, masks, mask_info, alpha=0.7):
     """
     Draw segmentation on image
     """
-    mask_color_id = 0
     w_ratio = 0.4
     color_list = get_colormap(rgb=True)
     im = np.array(im).astype("float32")

paddlex/inference/models/semantic_segmentation/__init__.py

@@ -1,4 +1,4 @@
-# copyright (c) 2024 PaddlePaddle Authors. All Rights Reserve.
+# Copyright (c) 2024 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.

paddlex/inference/models/semantic_segmentation/predictor.py

@@ -1,4 +1,4 @@
-# copyright (c) 2024 PaddlePaddle Authors. All Rights Reserve.
+# Copyright (c) 2024 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.
@@ -12,27 +12,22 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from typing import Any, Union, Dict, List, Tuple
+from typing import Any, Dict, List, Tuple, Union
+
 import numpy as np
 
-from ....utils.func_register import FuncRegister
 from ....modules.semantic_segmentation.model_list import MODELS
+from ....utils.func_register import FuncRegister
 from ...common.batch_sampler import ImageBatchSampler
 from ...common.reader import ReadImage
-from ..common import (
-    ResizeByShort,
-    Normalize,
-    ToCHWImage,
-    ToBatch,
-    StaticInfer,
-)
+from ..base import BasePredictor
+from ..common import Normalize, ToBatch, ToCHWImage
 from .processors import Resize, SegPostProcess
-from ..base import BasicPredictor
 from .result import SegResult
 
 
-class SegPredictor(BasicPredictor):
-    """SegPredictor that inherits from BasicPredictor."""
+class SegPredictor(BasePredictor):
+    """SegPredictor that inherits from BasePredictor."""
 
     entities = MODELS
 
@@ -95,11 +90,7 @@ class SegPredictor(BasicPredictor):
             _, op = self._FUNC_MAP["Resize"](self, target_size=self.target_size)
             preprocessors["Resize"] = op
 
-        infer = StaticInfer(
-            model_dir=self.model_dir,
-            model_prefix=self.MODEL_FILE_PREFIX,
-            option=self.pp_option,
-        )
+        infer = self.create_static_infer()
 
         postprocessers = SegPostProcess()
 

paddlex/inference/models/semantic_segmentation/processors.py

@@ -1,4 +1,4 @@
-# copyright (c) 2024 PaddlePaddle Authors. All Rights Reserve.
+# Copyright (c) 2024 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.
@@ -12,19 +12,20 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from typing import List, Tuple, Union
-import os
-import sys
-import cv2
-import copy
 import math
-import pyclipper
+
 import numpy as np
-from ..common.vision.processors import _BaseResize
 
+from ....utils.deps import class_requires_deps, is_dep_available
+from ...utils.benchmark import benchmark
 from ..common.vision import funcs as F
+from ..common.vision.processors import _BaseResize
+
+if is_dep_available("opencv-contrib-python"):
+    import cv2
 
 
+@benchmark.timeit
 class Resize(_BaseResize):
     """Resize the image."""
 
@@ -81,6 +82,8 @@ class Resize(_BaseResize):
         return img
 
 
+@benchmark.timeit
+@class_requires_deps("opencv-contrib-python")
 class SegPostProcess:
     """Semantic Segmentation PostProcess
 

paddlex/inference/models/semantic_segmentation/result.py

@@ -1,4 +1,4 @@
-# copyright (c) 2024 PaddlePaddle Authors. All Rights Reserve.
+# Copyright (c) 2024 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.
@@ -12,11 +12,12 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import copy
+
 import numpy as np
 from PIL import Image
-import copy
 
-from ...common.result import BaseCVResult, StrMixin, JsonMixin
+from ...common.result import BaseCVResult, JsonMixin
 
 
 class SegResult(BaseCVResult):

paddlex/inference/models/table_structure_recognition/__init__.py

@@ -1,4 +1,4 @@
-# copyright (c) 2024 PaddlePaddle Authors. All Rights Reserve.
+# Copyright (c) 2024 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.

paddlex/inference/models/table_structure_recognition/predictor.py

@@ -1,4 +1,4 @@
-# copyright (c) 2024 PaddlePaddle Authors. All Rights Reserve.
+# Copyright (c) 2024 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.
@@ -12,27 +12,21 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from typing import Any, Union, Dict, List, Tuple
+from typing import Any, Dict, List, Tuple, Union
+
 import numpy as np
 
-from ....utils.func_register import FuncRegister
 from ....modules.table_recognition.model_list import MODELS
+from ....utils.func_register import FuncRegister
 from ...common.batch_sampler import ImageBatchSampler
 from ...common.reader import ReadImage
-from ..common import (
-    Resize,
-    ResizeByLong,
-    Normalize,
-    ToCHWImage,
-    ToBatch,
-    StaticInfer,
-)
-from ..base import BasicPredictor
+from ..base import BasePredictor
+from ..common import Normalize, ResizeByLong, ToBatch, ToCHWImage
 from .processors import Pad, TableLabelDecode
 from .result import TableRecResult
 
 
-class TablePredictor(BasicPredictor):
+class TablePredictor(BasePredictor):
     entities = MODELS
 
     _FUNC_MAP = {}
@@ -59,11 +53,7 @@ class TablePredictor(BasicPredictor):
                 preprocessors.append(op)
         preprocessors.append(ToBatch())
 
-        infer = StaticInfer(
-            model_dir=self.model_dir,
-            model_prefix=self.MODEL_FILE_PREFIX,
-            option=self.pp_option,
-        )
+        infer = self.create_static_infer()
 
         postprocessors = TableLabelDecode(
             model_name=self.config["Global"]["model_name"],

paddlex/inference/models/table_structure_recognition/processors.py

@@ -1,4 +1,4 @@
-# copyright (c) 2024 PaddlePaddle Authors. All Rights Reserve.
+# Copyright (c) 2024 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.
@@ -13,12 +13,13 @@
 # limitations under the License.
 
 
-import cv2
 import numpy as np
-from numpy import ndarray
+
+from ...utils.benchmark import benchmark
 from ..common.vision import funcs as F
 
 
+@benchmark.timeit
 class Pad:
     """Pad the image."""
 
@@ -60,6 +61,7 @@ class Pad:
         return [self.apply(img) for img in imgs]
 
 
+@benchmark.timeit
 class TableLabelDecode:
     """decode the table model outputs(probs) to character str"""
 
@@ -121,16 +123,8 @@ class TableLabelDecode:
 
     def __call__(self, pred, img_size, ori_img_size):
         """apply"""
-        bbox_preds, structure_probs = [], []
-
-        for i in range(len(pred[0][0])):
-            bbox_preds.append(pred[0][0][i])
-            structure_probs.append(pred[1][0][i])
-        bbox_preds = [bbox_preds]
-        structure_probs = [structure_probs]
-
-        bbox_preds = np.array(bbox_preds)
-        structure_probs = np.array(structure_probs)
+        bbox_preds = np.array([list(pred[0][0])])
+        structure_probs = np.array([list(pred[1][0])])
 
         bbox_list, structure_str_list, structure_score = self.decode(
             structure_probs, bbox_preds, img_size, ori_img_size
@@ -159,9 +153,11 @@ class TableLabelDecode:
         structure_batch_list = []
         bbox_batch_list = []
         batch_size = len(structure_idx)
+        bbox_list = []
+        scale_list = []
+        scales = [0] * 8
         for batch_idx in range(batch_size):
             structure_list = []
-            bbox_list = []
             score_list = []
             for idx in range(len(structure_idx[batch_idx])):
                 char_idx = int(structure_idx[batch_idx][idx])
@@ -172,15 +168,21 @@ class TableLabelDecode:
                 text = self.character[char_idx]
                 if text in self.td_token:
                     bbox = bbox_preds[batch_idx, idx]
-                    bbox = self._bbox_decode(
-                        bbox, padding_size[batch_idx], ori_img_size[batch_idx]
+                    h_scale, w_scale = self._get_bbox_scales(
+                        padding_size[batch_idx], ori_img_size[batch_idx]
                     )
-                    bbox_list.append(bbox.astype(int))
+                    scales[0::2] = [h_scale] * 4
+                    scales[1::2] = [w_scale] * 4
+                    bbox_list.append(bbox)
+                    scale_list.append(scales)
+
                 structure_list.append(text)
                 score_list.append(structure_probs[batch_idx, idx])
             structure_batch_list.append(structure_list)
             structure_score = np.mean(score_list)
-            bbox_batch_list.append(bbox_list)
+
+        bbox_batch_array = np.multiply(np.array(bbox_list), np.array(scale_list))
+        bbox_batch_list = [bbox_batch_array.astype(int).tolist()]
 
         return bbox_batch_list, structure_batch_list, structure_score
 
@@ -214,22 +216,14 @@ class TableLabelDecode:
             bbox_batch_list.append(bbox_list)
         return bbox_batch_list, structure_batch_list
 
-    def _bbox_decode(self, bbox, padding_shape, ori_shape):
-
+    def _get_bbox_scales(self, padding_shape, ori_shape):
         if self.model_name == "SLANet":
             w, h = ori_shape
-            bbox[0::2] *= w
-            bbox[1::2] *= h
+            return w, h
         else:
             w, h = padding_shape
             ori_w, ori_h = ori_shape
             ratio_w = w / ori_w
             ratio_h = h / ori_h
             ratio = min(ratio_w, ratio_h)
-
-            bbox[0::2] *= w
-            bbox[1::2] *= h
-            bbox[0::2] /= ratio
-            bbox[1::2] /= ratio
-
-        return bbox
+            return w / ratio, h / ratio

paddlex/inference/models/table_structure_recognition/result.py

@@ -1,4 +1,4 @@
-# copyright (c) 2024 PaddlePaddle Authors. All Rights Reserve.
+# Copyright (c) 2024 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.
@@ -13,27 +13,20 @@
 # limitations under the License.
 
 import copy
-import cv2
+
 import numpy as np
-from pathlib import Path
-import PIL
-from PIL import Image, ImageDraw, ImageFont
 
-from ...common.result import BaseCVResult, StrMixin, JsonMixin
+from ....utils.deps import class_requires_deps, is_dep_available
+from ...common.result import BaseCVResult, JsonMixin
+
+if is_dep_available("opencv-contrib-python"):
+    import cv2
 
 
+@class_requires_deps("opencv-contrib-python")
 class TableRecResult(BaseCVResult):
     """SaveTableResults"""
 
-    def _get_input_fn(self):
-        fn = super()._get_input_fn()
-        if (page_idx := self["page_index"]) is not None:
-            fp = Path(fn)
-            stem, suffix = fp.stem, fp.suffix
-            return f"{stem}_{page_idx}{suffix}"
-        else:
-            return fn
-
     def _to_img(self):
         image = self["input_img"]
         bbox_res = self["bbox"]

paddlex/inference/models/text_detection/__init__.py

@@ -1,4 +1,4 @@
-# copyright (c) 2024 PaddlePaddle Authors. All Rights Reserve.
+# Copyright (c) 2024 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.

paddlex/inference/models/text_detection/predictor.py

@@ -1,4 +1,4 @@
-# copyright (c) 2024 PaddlePaddle Authors. All Rights Reserve.
+# Copyright (c) 2024 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.
@@ -12,27 +12,21 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import numpy as np
 from typing import List, Union
 
-from ....utils.func_register import FuncRegister
+import numpy as np
+
 from ....modules.text_detection.model_list import MODELS
+from ....utils.func_register import FuncRegister
 from ...common.batch_sampler import ImageBatchSampler
 from ...common.reader import ReadImage
-from ..common import (
-    Resize,
-    ResizeByShort,
-    Normalize,
-    ToCHWImage,
-    ToBatch,
-    StaticInfer,
-)
-from ..base import BasicPredictor
-from .processors import DetResizeForTest, NormalizeImage, DBPostProcess
+from ..base import BasePredictor
+from ..common import ToBatch, ToCHWImage
+from .processors import DBPostProcess, DetResizeForTest, NormalizeImage
 from .result import TextDetResult
 
 
-class TextDetPredictor(BasicPredictor):
+class TextDetPredictor(BasePredictor):
 
     entities = MODELS
 
@@ -46,6 +40,8 @@ class TextDetPredictor(BasicPredictor):
         thresh: Union[float, None] = None,
         box_thresh: Union[float, None] = None,
         unclip_ratio: Union[float, None] = None,
+        input_shape=None,
+        max_side_limit: int = 4000,
         *args,
         **kwargs
     ):
@@ -56,6 +52,8 @@ class TextDetPredictor(BasicPredictor):
         self.thresh = thresh
         self.box_thresh = box_thresh
         self.unclip_ratio = unclip_ratio
+        self.input_shape = input_shape
+        self.max_side_limit = max_side_limit
         self.pre_tfs, self.infer, self.post_op = self._build()
 
     def _build_batch_sampler(self):
@@ -76,11 +74,7 @@ class TextDetPredictor(BasicPredictor):
                 pre_tfs[name] = op
         pre_tfs["ToBatch"] = ToBatch()
 
-        infer = StaticInfer(
-            model_dir=self.model_dir,
-            model_prefix=self.MODEL_FILE_PREFIX,
-            option=self.pp_option,
-        )
+        infer = self.create_static_infer()
 
         post_op = self.build_postprocess(**self.config["PostProcess"])
         return pre_tfs, infer, post_op
@@ -93,6 +87,7 @@ class TextDetPredictor(BasicPredictor):
         thresh: Union[float, None] = None,
         box_thresh: Union[float, None] = None,
         unclip_ratio: Union[float, None] = None,
+        max_side_limit: Union[int, None] = None,
     ):
 
         batch_raw_imgs = self.pre_tfs["Read"](imgs=batch_data.instances)
@@ -100,6 +95,9 @@ class TextDetPredictor(BasicPredictor):
             imgs=batch_raw_imgs,
             limit_side_len=limit_side_len or self.limit_side_len,
             limit_type=limit_type or self.limit_type,
+            max_side_limit=(
+                max_side_limit if max_side_limit is not None else self.max_side_limit
+            ),
         )
         batch_imgs = self.pre_tfs["Normalize"](imgs=batch_imgs)
         batch_imgs = self.pre_tfs["ToCHW"](imgs=batch_imgs)
@@ -135,6 +133,8 @@ class TextDetPredictor(BasicPredictor):
         # TODO: align to PaddleOCR
 
         if self.model_name in (
+            "PP-OCRv5_server_det",
+            "PP-OCRv5_mobile_det",
             "PP-OCRv4_server_det",
             "PP-OCRv4_mobile_det",
             "PP-OCRv3_server_det",
@@ -147,7 +147,10 @@ class TextDetPredictor(BasicPredictor):
             limit_type = self.limit_type or kwargs.get("limit_type", "min")
 
         return "Resize", DetResizeForTest(
-            limit_side_len=limit_side_len, limit_type=limit_type, **kwargs
+            limit_side_len=limit_side_len,
+            limit_type=limit_type,
+            input_shape=self.input_shape,
+            **kwargs
         )
 
     @register("NormalizeImage")
@@ -157,11 +160,8 @@ class TextDetPredictor(BasicPredictor):
         std=[0.229, 0.224, 0.225],
         scale=1 / 255,
         order="",
-        channel_num=3,
     ):
-        return "Normalize", NormalizeImage(
-            mean=mean, std=std, scale=scale, order=order, channel_num=channel_num
-        )
+        return "Normalize", NormalizeImage(mean=mean, std=std, scale=scale, order=order)
 
     @register("ToCHWImage")
     def build_to_chw(self):