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

paddlex/inference/pipelines/table_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.
@@ -12,14 +12,14 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import os
-from typing import Dict
+import copy
 from pathlib import Path
-from PIL import Image, ImageDraw
+from typing import Dict
+
 import numpy as np
-import cv2
-import copy
-from ...common.result import BaseCVResult, HtmlMixin, XlsxMixin, StrMixin, JsonMixin
+from PIL import Image, ImageDraw
+
+from ...common.result import BaseCVResult, HtmlMixin, JsonMixin, XlsxMixin
 
 
 class SingleTableRecognitionResult(BaseCVResult, HtmlMixin, XlsxMixin):
@@ -96,7 +96,7 @@ class TableRecognitionResult(BaseCVResult, HtmlMixin, XlsxMixin):
         super().__init__(data)
         HtmlMixin.__init__(self)
         XlsxMixin.__init__(self)
-    
+
     def _get_input_fn(self):
         fn = super()._get_input_fn()
         if (page_idx := self["page_index"]) is not None:
@@ -120,7 +120,7 @@ class TableRecognitionResult(BaseCVResult, HtmlMixin, XlsxMixin):
 
         if len(self["table_res_list"]) > 0:
             table_cell_img = Image.fromarray(
-                copy.deepcopy(self["doc_preprocessor_res"]["output_img"])
+                copy.deepcopy(self["doc_preprocessor_res"]["output_img"][:, :, ::-1])
             )
             table_draw = ImageDraw.Draw(table_cell_img)
             rectangle_color = (255, 0, 0)
@@ -198,7 +198,9 @@ class TableRecognitionResult(BaseCVResult, HtmlMixin, XlsxMixin):
             table_region_id = table_res["table_region_id"]
             key = f"table_{table_region_id}"
             res_html_dict[key] = table_res.html["pred"]
-            res_html_dict[key] = res_html_dict[key].replace("<table>", '<table border="1">')
+            res_html_dict[key] = res_html_dict[key].replace(
+                "<table>", '<table border="1">'
+            )
         return res_html_dict
 
     def _to_xlsx(self) -> Dict[str, str]:

paddlex/inference/pipelines/table_recognition/table_recognition_post_processing.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.
@@ -11,12 +11,14 @@
 # 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 typing import Any, Dict, Optional
+from typing import Dict
+
 import numpy as np
-from ..layout_parsing.utils import get_sub_regions_ocr_res
+
 from ..components import convert_points_to_boxes
-from .result import SingleTableRecognitionResult
+from ..layout_parsing.utils import get_sub_regions_ocr_res
 from ..ocr.result import OCRResult
+from .result import SingleTableRecognitionResult
 
 
 def get_ori_image_coordinate(x: int, y: int, box_list: list) -> list:
@@ -323,10 +325,12 @@ def get_table_recognition_res(
     crop_start_point = [table_box[0][0], table_box[0][1]]
     img_shape = overall_ocr_res["doc_preprocessor_res"]["output_img"].shape[0:2]
 
-    if len(table_structure_pred['bbox']) == 0 or len(table_ocr_pred["rec_boxes"]) == 0:
-        pred_html = ' '.join(list(table_structure_pred["structure"]))
-        if len(table_structure_pred['bbox']) != 0:
-            convert_table_structure_pred_bbox(table_structure_pred, crop_start_point, img_shape)
+    if len(table_structure_pred["bbox"]) == 0 or len(table_ocr_pred["rec_boxes"]) == 0:
+        pred_html = " ".join(list(table_structure_pred["structure"]))
+        if len(table_structure_pred["bbox"]) != 0:
+            convert_table_structure_pred_bbox(
+                table_structure_pred, crop_start_point, img_shape
+            )
             table_cells_result = table_structure_pred["cell_box_list"]
         else:
             table_cells_result = []

paddlex/inference/pipelines/table_recognition/table_recognition_post_processing_v2.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.
@@ -12,12 +12,13 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import math
-from typing import Any, Dict, Optional
+
 import numpy as np
-from ..layout_parsing.utils import get_sub_regions_ocr_res
+
 from ..components import convert_points_to_boxes
-from .result import SingleTableRecognitionResult
+from ..layout_parsing.utils import get_sub_regions_ocr_res
 from ..ocr.result import OCRResult
+from .result import SingleTableRecognitionResult
 
 
 def get_ori_image_coordinate(x: int, y: int, box_list: list) -> list:
@@ -130,8 +131,8 @@ def compute_inter(rec1, rec2):
     Returns:
         float: Intersection over rec2_area
     """
-    x1_1, y1_1, x2_1, y2_1 = rec1
-    x1_2, y1_2, x2_2, y2_2 = rec2
+    x1_1, y1_1, x2_1, y2_1 = map(float, rec1)
+    x1_2, y1_2, x2_2, y2_2 = map(float, rec2)
     x_left = max(x1_1, x1_2)
     y_top = max(y1_1, y1_2)
     x_right = min(x2_1, x2_2)
@@ -141,7 +142,7 @@ def compute_inter(rec1, rec2):
     inter_area = inter_width * inter_height
     rec2_area = (x2_2 - x1_2) * (y2_2 - y1_2)
     if rec2_area == 0:
-        return 0 
+        return 0
     iou = inter_area / rec2_area
     return iou
 
@@ -158,9 +159,11 @@ def match_table_and_ocr(cell_box_list, ocr_dt_boxes, table_cells_flag, row_start
         dict: matched dict, key is table index, value is ocr index
     """
     all_matched = []
-    for k in range(len(table_cells_flag)-1):
+    for k in range(len(table_cells_flag) - 1):
         matched = {}
-        for i, table_box in enumerate(cell_box_list[table_cells_flag[k]:table_cells_flag[k+1]]):
+        for i, table_box in enumerate(
+            cell_box_list[table_cells_flag[k] : table_cells_flag[k + 1]]
+        ):
             if len(table_box) == 8:
                 table_box = [
                     np.min(table_box[0::2]),
@@ -174,13 +177,13 @@ def match_table_and_ocr(cell_box_list, ocr_dt_boxes, table_cells_flag, row_start
                         matched[i] = [j]
                     else:
                         matched[i].append(j)
-        real_len=max(matched.keys())+1 if len(matched)!=0 else 0
-        if table_cells_flag[k+1] < row_start_index[k+1]:
-            for s in range(row_start_index[k+1]-table_cells_flag[k+1]):
-                matched[real_len+s] = []
-        elif table_cells_flag[k+1] > row_start_index[k+1]:
-            for s in range(table_cells_flag[k+1]-row_start_index[k+1]):
-                matched[real_len-1].append(matched[real_len+s])
+        real_len = max(matched.keys()) + 1 if len(matched) != 0 else 0
+        if table_cells_flag[k + 1] < row_start_index[k + 1]:
+            for s in range(row_start_index[k + 1] - table_cells_flag[k + 1]):
+                matched[real_len + s] = []
+        elif table_cells_flag[k + 1] > row_start_index[k + 1]:
+            for s in range(table_cells_flag[k + 1] - row_start_index[k + 1]):
+                matched[real_len - 1].append(matched[real_len + s])
         all_matched.append(matched)
     return all_matched
 
@@ -212,7 +215,7 @@ def get_html_result(
             if "<td></td>" == tag:
                 pred_html.extend("<td>")
             if td_index in matched_index.keys():
-                if len(matched_index[td_index])==0:
+                if len(matched_index[td_index]) == 0:
                     continue
                 b_with = False
                 if (
@@ -245,7 +248,10 @@ def get_html_result(
                 pred_html.append(tag)
             td_index += 1
             td_count += 1
-            if td_count>=table_cells_flag[matched_list_index+1] and matched_list_index<len(all_matched_index)-1:
+            if (
+                td_count >= table_cells_flag[matched_list_index + 1]
+                and matched_list_index < len(all_matched_index) - 1
+            ):
                 matched_list_index += 1
                 td_index = 0
         else:
@@ -299,15 +305,15 @@ def sort_table_cells_boxes(boxes):
 
 def convert_to_four_point_coordinates(boxes):
     """
-    Convert bounding boxes from [x1, y1, x2, y2] format to 
+    Convert bounding boxes from [x1, y1, x2, y2] format to
     [x1, y1, x2, y1, x2, y2, x1, y2] format.
 
     Parameters:
-    - boxes: A list of bounding boxes, each defined as a list of integers 
+    - boxes: A list of bounding boxes, each defined as a list of integers
              in the format [x1, y1, x2, y2].
 
     Returns:
-    - A list of bounding boxes, each converted to the format 
+    - A list of bounding boxes, each converted to the format
       [x1, y1, x2, y1, x2, y2, x1, y2].
     """
     # Initialize an empty list to store the converted bounding boxes
@@ -316,7 +322,7 @@ def convert_to_four_point_coordinates(boxes):
     # Loop over each box in the input list
     for box in boxes:
         x1, y1, x2, y2 = box
-        
+
         # Define the four corner points
         top_left = (x1, y1)
         top_right = (x2, y1)
@@ -325,10 +331,14 @@ def convert_to_four_point_coordinates(boxes):
 
         # Create a new list for the converted box
         converted_box = [
-            top_left[0], top_left[1],  # Top-left corner
-            top_right[0], top_right[1],  # Top-right corner
-            bottom_right[0], bottom_right[1],  # Bottom-right corner
-            bottom_left[0], bottom_left[1]   # Bottom-left corner
+            top_left[0],
+            top_left[1],  # Top-left corner
+            top_right[0],
+            top_right[1],  # Top-right corner
+            bottom_right[0],
+            bottom_right[1],  # Bottom-right corner
+            bottom_left[0],
+            bottom_left[1],  # Bottom-left corner
         ]
 
         # Append the converted box to the list
@@ -339,13 +349,13 @@ def convert_to_four_point_coordinates(boxes):
 
 def find_row_start_index(html_list):
     """
-        find the index of the first cell in each row
+    find the index of the first cell in each row
 
-        Args:
-            html_list (list): list for html results
+    Args:
+        html_list (list): list for html results
 
-        Returns:
-            row_start_indices (list): list for the index of the first cell in each row
+    Returns:
+        row_start_indices (list): list for the index of the first cell in each row
     """
     # Initialize an empty list to store the indices of row start positions
     row_start_indices = []
@@ -403,8 +413,10 @@ def get_table_recognition_res(
     table_structure_result: list,
     table_cells_result: list,
     overall_ocr_res: OCRResult,
+    table_ocr_pred: dict,
     cells_texts_list: list,
     use_table_cells_ocr_results: bool,
+    use_table_cells_split_ocr: bool,
 ) -> SingleTableRecognitionResult:
     """
     Retrieve table recognition result from cropped image info, table structure prediction, and overall OCR result.
@@ -414,6 +426,7 @@ def get_table_recognition_res(
         table_structure_result (list): Predicted table structure.
         table_cells_result (list): Predicted table cells.
         overall_ocr_res (OCRResult): Overall OCR result from the input image.
+        table_ocr_pred (dict): Table OCR result from the input image.
         cells_texts_list (list): OCR results with cells.
         use_table_cells_ocr_results (bool): whether to use OCR results with cells.
 
@@ -422,15 +435,16 @@ def get_table_recognition_res(
     """
 
     table_cells_result = convert_to_four_point_coordinates(table_cells_result)
-
     table_box = np.array([table_box])
-    table_ocr_pred = get_sub_regions_ocr_res(overall_ocr_res, table_box)
+
+    if not (use_table_cells_ocr_results == True and use_table_cells_split_ocr == True):
+        table_ocr_pred = get_sub_regions_ocr_res(overall_ocr_res, table_box)
 
     crop_start_point = [table_box[0][0], table_box[0][1]]
     img_shape = overall_ocr_res["doc_preprocessor_res"]["output_img"].shape[0:2]
 
     if len(table_cells_result) == 0 or len(table_ocr_pred["rec_boxes"]) == 0:
-        pred_html = ' '.join(table_structure_result)
+        pred_html = " ".join(table_structure_result)
         if len(table_cells_result) != 0:
             table_cells_result = convert_table_structure_pred_bbox(
                 table_cells_result, crop_start_point, img_shape
@@ -446,7 +460,7 @@ def get_table_recognition_res(
         table_cells_result, crop_start_point, img_shape
     )
 
-    if use_table_cells_ocr_results == True:
+    if use_table_cells_ocr_results == True and use_table_cells_split_ocr == False:
         ocr_dt_boxes = table_cells_result
         ocr_texts_res = cells_texts_list
     else:
@@ -458,8 +472,12 @@ def get_table_recognition_res(
     table_cells_flag = map_and_get_max(table_cells_flag, row_start_index)
     table_cells_flag.append(len(table_cells_result))
     row_start_index.append(len(table_cells_result))
-    matched_index = match_table_and_ocr(table_cells_result, ocr_dt_boxes, table_cells_flag, table_cells_flag)
-    pred_html = get_html_result(matched_index, ocr_texts_res, table_structure_result, row_start_index)
+    matched_index = match_table_and_ocr(
+        table_cells_result, ocr_dt_boxes, table_cells_flag, table_cells_flag
+    )
+    pred_html = get_html_result(
+        matched_index, ocr_texts_res, table_structure_result, row_start_index
+    )
 
     single_img_res = {
         "cell_box_list": table_cells_result,

paddlex/inference/pipelines/table_recognition/utils.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/pipelines/ts_anomaly_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/pipelines/ts_anomaly_detection/pipeline.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.
@@ -12,15 +12,18 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from typing import Any, Dict, Optional, Union, List
+from typing import Any, Dict, List, Optional, Union
+
 import pandas as pd
 
+from ....utils.deps import pipeline_requires_extra
+from ...models.ts_anomaly_detection.result import TSAdResult
+from ...utils.hpi import HPIConfig
 from ...utils.pp_option import PaddlePredictorOption
 from ..base import BasePipeline
 
-from ...models.ts_anomaly_detection.result import TSAdResult
-
 
+@pipeline_requires_extra("ts")
 class TSAnomalyDetPipeline(BasePipeline):
     """TSAnomalyDetPipeline Pipeline"""
 
@@ -32,6 +35,7 @@ class TSAnomalyDetPipeline(BasePipeline):
         device: str = None,
         pp_option: PaddlePredictorOption = None,
         use_hpip: bool = False,
+        hpi_config: Optional[Union[Dict[str, Any], HPIConfig]] = None,
     ) -> None:
         """Initializes the Time Series ad pipeline.
 
@@ -39,10 +43,16 @@ class TSAnomalyDetPipeline(BasePipeline):
             config (Dict): Configuration dictionary containing various settings.
             device (str, optional): Device to run the predictions on. Defaults to None.
             pp_option (PaddlePredictorOption, optional): PaddlePredictor options. Defaults to None.
-            use_hpip (bool, optional): Whether to use high-performance inference (hpip) for prediction. Defaults to False.
+            use_hpip (bool, optional): Whether to use the high-performance
+                inference plugin (HPIP) by default. Defaults to False.
+            hpi_config (Optional[Union[Dict[str, Any], HPIConfig]], optional):
+                The default high-performance inference configuration dictionary.
+                Defaults to None.
         """
 
-        super().__init__(device=device, pp_option=pp_option, use_hpip=use_hpip)
+        super().__init__(
+            device=device, pp_option=pp_option, use_hpip=use_hpip, hpi_config=hpi_config
+        )
 
         ts_ad_model_config = config["SubModules"]["TSAnomalyDetection"]
         self.ts_ad_model = self.create_model(ts_ad_model_config)

paddlex/inference/pipelines/ts_classification/__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/pipelines/ts_classification/pipeline.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.
@@ -12,15 +12,18 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from typing import Any, Dict, Optional, Union, List
+from typing import Any, Dict, List, Optional, Union
+
 import pandas as pd
 
+from ....utils.deps import pipeline_requires_extra
+from ...models.ts_classification.result import TSClsResult
+from ...utils.hpi import HPIConfig
 from ...utils.pp_option import PaddlePredictorOption
 from ..base import BasePipeline
 
-from ...models.ts_classification.result import TSClsResult
-
 
+@pipeline_requires_extra("ts")
 class TSClsPipeline(BasePipeline):
     """TSClsPipeline Pipeline"""
 
@@ -32,6 +35,7 @@ class TSClsPipeline(BasePipeline):
         device: str = None,
         pp_option: PaddlePredictorOption = None,
         use_hpip: bool = False,
+        hpi_config: Optional[Union[Dict[str, Any], HPIConfig]] = None,
     ) -> None:
         """Initializes the Time Series classification pipeline.
 
@@ -39,10 +43,16 @@ class TSClsPipeline(BasePipeline):
             config (Dict): Configuration dictionary containing various settings.
             device (str, optional): Device to run the predictions on. Defaults to None.
             pp_option (PaddlePredictorOption, optional): PaddlePredictor options. Defaults to None.
-            use_hpip (bool, optional): Whether to use high-performance inference (hpip) for prediction. Defaults to False.
+            use_hpip (bool, optional): Whether to use the high-performance
+                inference plugin (HPIP) by default. Defaults to False.
+            hpi_config (Optional[Union[Dict[str, Any], HPIConfig]], optional):
+                The default high-performance inference configuration dictionary.
+                Defaults to None.
         """
 
-        super().__init__(device=device, pp_option=pp_option, use_hpip=use_hpip)
+        super().__init__(
+            device=device, pp_option=pp_option, use_hpip=use_hpip, hpi_config=hpi_config
+        )
 
         ts_classification_model_config = config["SubModules"]["TSClassification"]
         self.ts_classification_model = self.create_model(ts_classification_model_config)

paddlex/inference/pipelines/ts_forecasting/__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/pipelines/ts_forecasting/pipeline.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.
@@ -12,15 +12,18 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from typing import Any, Dict, Optional, Union, List
+from typing import Any, Dict, List, Optional, Union
+
 import pandas as pd
 
+from ....utils.deps import pipeline_requires_extra
+from ...models.ts_forecasting.result import TSFcResult
+from ...utils.hpi import HPIConfig
 from ...utils.pp_option import PaddlePredictorOption
 from ..base import BasePipeline
 
-from ...models.ts_forecasting.result import TSFcResult
-
 
+@pipeline_requires_extra("ts")
 class TSFcPipeline(BasePipeline):
     """TSFcPipeline Pipeline"""
 
@@ -32,6 +35,7 @@ class TSFcPipeline(BasePipeline):
         device: str = None,
         pp_option: PaddlePredictorOption = None,
         use_hpip: bool = False,
+        hpi_config: Optional[Union[Dict[str, Any], HPIConfig]] = None,
     ) -> None:
         """Initializes the Time Series Forecast pipeline.
 
@@ -39,10 +43,16 @@ class TSFcPipeline(BasePipeline):
             config (Dict): Configuration dictionary containing various settings.
             device (str, optional): Device to run the predictions on. Defaults to None.
             pp_option (PaddlePredictorOption, optional): PaddlePredictor options. Defaults to None.
-            use_hpip (bool, optional): Whether to use high-performance inference (hpip) for prediction. Defaults to False.
+            use_hpip (bool, optional): Whether to use the high-performance
+                inference plugin (HPIP) by default. Defaults to False.
+            hpi_config (Optional[Union[Dict[str, Any], HPIConfig]], optional):
+                The default high-performance inference configuration dictionary.
+                Defaults to None.
         """
 
-        super().__init__(device=device, pp_option=pp_option, use_hpip=use_hpip)
+        super().__init__(
+            device=device, pp_option=pp_option, use_hpip=use_hpip, hpi_config=hpi_config
+        )
 
         ts_forecast_model_config = config["SubModules"]["TSForecast"]
         self.ts_forecast_model = self.create_model(ts_forecast_model_config)

paddlex/inference/pipelines/video_classification/__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/pipelines/video_classification/pipeline.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.
@@ -12,14 +12,18 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from typing import Any, Dict, Optional, Union, List
+from typing import Any, Dict, List, Optional, Union
+
 import numpy as np
-from ...utils.pp_option import PaddlePredictorOption
-from ..base import BasePipeline
 
+from ....utils.deps import pipeline_requires_extra
 from ...models.video_classification.result import TopkVideoResult
+from ...utils.hpi import HPIConfig
+from ...utils.pp_option import PaddlePredictorOption
+from ..base import BasePipeline
 
 
+@pipeline_requires_extra("video")
 class VideoClassificationPipeline(BasePipeline):
     """Video Classification Pipeline"""
 
@@ -31,6 +35,7 @@ class VideoClassificationPipeline(BasePipeline):
         device: str = None,
         pp_option: PaddlePredictorOption = None,
         use_hpip: bool = False,
+        hpi_config: Optional[Union[Dict[str, Any], HPIConfig]] = None,
     ) -> None:
         """
         Initializes the class with given configurations and options.
@@ -39,9 +44,15 @@ class VideoClassificationPipeline(BasePipeline):
             config (Dict): Configuration dictionary containing model and other parameters.
             device (str): The device to run the prediction on. Default is None.
             pp_option (PaddlePredictorOption): Options for PaddlePaddle predictor. Default is None.
-            use_hpip (bool): Whether to use high-performance inference (hpip) for prediction. Defaults to False.
+            use_hpip (bool, optional): Whether to use the high-performance
+                inference plugin (HPIP) by default. Defaults to False.
+            hpi_config (Optional[Union[Dict[str, Any], HPIConfig]], optional):
+                The default high-performance inference configuration dictionary.
+                Defaults to None.
         """
-        super().__init__(device=device, pp_option=pp_option, use_hpip=use_hpip)
+        super().__init__(
+            device=device, pp_option=pp_option, use_hpip=use_hpip, hpi_config=hpi_config
+        )
 
         video_classification_model_config = config["SubModules"]["VideoClassification"]
         self.video_classification_model = self.create_model(

paddlex/inference/pipelines/video_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/pipelines/video_detection/pipeline.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.
@@ -12,14 +12,18 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from typing import Any, Dict, Optional, Union, List
+from typing import Any, Dict, List, Optional, Union
+
 import numpy as np
-from ...utils.pp_option import PaddlePredictorOption
-from ..base import BasePipeline
 
+from ....utils.deps import pipeline_requires_extra
 from ...models.video_detection.result import DetVideoResult
+from ...utils.hpi import HPIConfig
+from ...utils.pp_option import PaddlePredictorOption
+from ..base import BasePipeline
 
 
+@pipeline_requires_extra("video")
 class VideoDetectionPipeline(BasePipeline):
     """Video detection Pipeline"""
 
@@ -31,6 +35,7 @@ class VideoDetectionPipeline(BasePipeline):
         device: str = None,
         pp_option: PaddlePredictorOption = None,
         use_hpip: bool = False,
+        hpi_config: Optional[Union[Dict[str, Any], HPIConfig]] = None,
     ) -> None:
         """
         Initializes the class with given configurations and options.
@@ -39,9 +44,15 @@ class VideoDetectionPipeline(BasePipeline):
             config (Dict): Configuration dictionary containing model and other parameters.
             device (str): The device to run the prediction on. Default is None.
             pp_option (PaddlePredictorOption): Options for PaddlePaddle predictor. Default is None.
-            use_hpip (bool): Whether to use high-performance inference (hpip) for prediction. Defaults to False.
+            use_hpip (bool, optional): Whether to use the high-performance
+                inference plugin (HPIP) by default. Defaults to False.
+            hpi_config (Optional[Union[Dict[str, Any], HPIConfig]], optional):
+                The default high-performance inference configuration dictionary.
+                Defaults to None.
         """
-        super().__init__(device=device, pp_option=pp_option, use_hpip=use_hpip)
+        super().__init__(
+            device=device, pp_option=pp_option, use_hpip=use_hpip, hpi_config=hpi_config
+        )
 
         video_detection_model_config = config["SubModules"]["VideoDetection"]
         model_kwargs = {}
@@ -49,7 +60,9 @@ class VideoDetectionPipeline(BasePipeline):
             model_kwargs["nms_thresh"] = video_detection_model_config["nms_thresh"]
         if "score_thresh" in video_detection_model_config:
             model_kwargs["score_thresh"] = video_detection_model_config["score_thresh"]
-        self.video_detection_model = self.create_model(video_detection_model_config, **model_kwargs)
+        self.video_detection_model = self.create_model(
+            video_detection_model_config, **model_kwargs
+        )
 
     def predict(
         self,

paddlex/inference/serving/__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.
@@ -11,3 +11,7 @@
 # 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 ...utils.deps import require_serving_plugin
+
+require_serving_plugin()