paddlex 3.0.2__py3-none-any.whl → 3.1.0__py3-none-any.whl

paddlex/inference/serving/basic_serving/_pipeline_apps/pp_chatocrv4_doc.py

@@ -46,7 +46,9 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
         pipeline = ctx.pipeline
 
         log_id = serving_utils.generate_log_id()
-
+        visualize_enabled = (
+            request.visualize if request.visualize is not None else ctx.config.visualize
+        )
         images, data_info = await ocr_common.get_images(request, ctx)
 
         result = await pipeline.call(
@@ -79,7 +81,7 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
         visual_info: List[dict] = []
         for i, (img, item) in enumerate(zip(images, result)):
             pruned_res = common.prune_result(item["layout_parsing_result"].json["res"])
-            if ctx.config.visualize:
+            if visualize_enabled:
                 imgs = {
                     "input_img": img,
                     **item["layout_parsing_result"].img,
@@ -127,19 +129,13 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
     ) -> AIStudioResultResponse[schema.BuildVectorStoreResult]:
         pipeline = ctx.pipeline
 
-        kwargs: Dict[str, Any] = {
-            "flag_save_bytes_vector": True,
-            "retriever_config": request.retrieverConfig,
-        }
-        if request.minCharacters is not None:
-            kwargs["min_characters"] = request.minCharacters
-        if request.blockSize is not None:
-            kwargs["block_size"] = request.blockSize
-
         vector_info = await serving_utils.call_async(
             pipeline.pipeline.build_vector,
             request.visualInfo,
-            **kwargs,
+            min_characters=request.minCharacters,
+            block_size=request.blockSize,
+            flag_save_bytes_vector=True,
+            retriever_config=request.retrieverConfig,
         )
 
         return AIStudioResultResponse[schema.BuildVectorStoreResult](
@@ -185,8 +181,13 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
     ) -> AIStudioResultResponse[schema.ChatResult]:
         pipeline = ctx.pipeline
 
-        kwargs: Dict[str, Any] = dict(
+        result = await serving_utils.call_async(
+            pipeline.pipeline.chat,
+            request.keyList,
+            request.visualInfo,
+            use_vector_retrieval=request.useVectorRetrieval,
             vector_info=request.vectorInfo,
+            min_characters=request.minCharacters,
             text_task_description=request.textTaskDescription,
             text_output_format=request.textOutputFormat,
             text_rules_str=request.textRulesStr,
@@ -197,22 +198,11 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
             table_rules_str=request.tableRulesStr,
             table_few_shot_demo_text_content=request.tableFewShotDemoTextContent,
             table_few_shot_demo_key_value_list=request.tableFewShotDemoKeyValueList,
+            mllm_predict_info=request.mllmPredictInfo,
+            mllm_integration_strategy=request.mllmIntegrationStrategy,
             chat_bot_config=request.chatBotConfig,
             retriever_config=request.retrieverConfig,
         )
-        if request.useVectorRetrieval is not None:
-            kwargs["use_vector_retrieval"] = request.useVectorRetrieval
-        if request.minCharacters is not None:
-            kwargs["min_characters"] = request.minCharacters
-        if request.mllmIntegrationStrategy is not None:
-            kwargs["mllm_integration_strategy"] = request.mllmIntegrationStrategy
-
-        result = await serving_utils.call_async(
-            pipeline.pipeline.chat,
-            request.keyList,
-            request.visualInfo,
-            **kwargs,
-        )
 
         return AIStudioResultResponse[schema.ChatResult](
             logId=serving_utils.generate_log_id(),

paddlex/inference/serving/basic_serving/_pipeline_apps/pp_doctranslation.py

@@ -0,0 +1,203 @@
+# Copyright (c) 2025 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.
+
+from typing import Any, Dict, List
+
+from .....utils.deps import function_requires_deps, is_dep_available
+from ...infra import utils as serving_utils
+from ...infra.config import AppConfig
+from ...infra.models import AIStudioResultResponse
+from ...schemas import pp_doctranslation as schema
+from .._app import create_app, primary_operation
+from ._common import common
+from ._common import ocr as ocr_common
+
+if is_dep_available("fastapi"):
+    from fastapi import FastAPI
+
+
+@function_requires_deps("fastapi")
+def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
+    pipeline.inintial_visual_predictor(pipeline.config)
+
+    app, ctx = create_app(
+        pipeline=pipeline, app_config=app_config, app_aiohttp_session=True
+    )
+
+    ocr_common.update_app_context(ctx)
+
+    @primary_operation(
+        app,
+        schema.ANALYZE_IMAGES_ENDPOINT,
+        "analyzeImages",
+    )
+    async def _analyze_images(
+        request: schema.AnalyzeImagesRequest,
+    ) -> AIStudioResultResponse[schema.AnalyzeImagesResult]:
+        pipeline = ctx.pipeline
+
+        log_id = serving_utils.generate_log_id()
+        visualize_enabled = (
+            request.visualize if request.visualize is not None else ctx.config.visualize
+        )
+        images, data_info = await ocr_common.get_images(request, ctx)
+
+        result = await pipeline.call(
+            pipeline.pipeline.visual_predict,
+            images,
+            use_doc_orientation_classify=request.useDocOrientationClassify,
+            use_doc_unwarping=request.useDocUnwarping,
+            use_textline_orientation=request.useTextlineOrientation,
+            use_seal_recognition=request.useSealRecognition,
+            use_table_recognition=request.useTableRecognition,
+            use_formula_recognition=request.useFormulaRecognition,
+            use_chart_recognition=request.useChartRecognition,
+            use_region_detection=request.useRegionDetection,
+            layout_threshold=request.layoutThreshold,
+            layout_nms=request.layoutNms,
+            layout_unclip_ratio=request.layoutUnclipRatio,
+            layout_merge_bboxes_mode=request.layoutMergeBboxesMode,
+            text_det_limit_side_len=request.textDetLimitSideLen,
+            text_det_limit_type=request.textDetLimitType,
+            text_det_thresh=request.textDetThresh,
+            text_det_box_thresh=request.textDetBoxThresh,
+            text_det_unclip_ratio=request.textDetUnclipRatio,
+            text_rec_score_thresh=request.textRecScoreThresh,
+            seal_det_limit_side_len=request.sealDetLimitSideLen,
+            seal_det_limit_type=request.sealDetLimitType,
+            seal_det_thresh=request.sealDetThresh,
+            seal_det_box_thresh=request.sealDetBoxThresh,
+            seal_det_unclip_ratio=request.sealDetUnclipRatio,
+            seal_rec_score_thresh=request.sealRecScoreThresh,
+            use_wired_table_cells_trans_to_html=request.useWiredTableCellsTransToHtml,
+            use_wireless_table_cells_trans_to_html=request.useWirelessTableCellsTransToHtml,
+            use_table_orientation_classify=request.useTableOrientationClassify,
+            use_ocr_results_with_table_cells=request.useOcrResultsWithTableCells,
+            use_e2e_wired_table_rec_model=request.useE2eWiredTableRecModel,
+            use_e2e_wireless_table_rec_model=request.useE2eWirelessTableRecModel,
+        )
+
+        layout_parsing_results: List[Dict[str, Any]] = []
+        for i, (img, item) in enumerate(zip(images, result)):
+            pruned_res = common.prune_result(item["layout_parsing_result"].json["res"])
+            md_data = item["layout_parsing_result"].markdown
+            md_text = md_data["markdown_texts"]
+            md_imgs = await serving_utils.call_async(
+                common.postprocess_images,
+                md_data["markdown_images"],
+                log_id,
+                filename_template=f"markdown_{i}/{{key}}",
+                file_storage=ctx.extra["file_storage"],
+                return_urls=ctx.extra["return_img_urls"],
+                max_img_size=ctx.extra["max_output_img_size"],
+            )
+            md_flags = md_data["page_continuation_flags"]
+            if visualize_enabled:
+                imgs = {
+                    "input_img": img,
+                    **item["layout_parsing_result"].img,
+                }
+                imgs = await serving_utils.call_async(
+                    common.postprocess_images,
+                    imgs,
+                    log_id,
+                    filename_template=f"{{key}}_{i}.jpg",
+                    file_storage=ctx.extra["file_storage"],
+                    return_urls=ctx.extra["return_img_urls"],
+                    max_img_size=ctx.extra["max_output_img_size"],
+                )
+            else:
+                imgs = {}
+            layout_parsing_results.append(
+                dict(
+                    prunedResult=pruned_res,
+                    markdown=dict(
+                        text=md_text,
+                        images=md_imgs,
+                        isStart=md_flags[0],
+                        isEnd=md_flags[1],
+                    ),
+                    outputImages=(
+                        {k: v for k, v in imgs.items() if k != "input_img"}
+                        if imgs
+                        else None
+                    ),
+                    inputImage=imgs.get("input_img"),
+                )
+            )
+
+        return AIStudioResultResponse[schema.AnalyzeImagesResult](
+            logId=log_id,
+            result=schema.AnalyzeImagesResult(
+                layoutParsingResults=layout_parsing_results,
+                dataInfo=data_info,
+            ),
+        )
+
+    @primary_operation(
+        app,
+        schema.TRANSLATE_ENDPOINT,
+        "translate",
+    )
+    async def _translate(
+        request: schema.TranslateRequest,
+    ) -> AIStudioResultResponse[schema.TranslateResult]:
+        pipeline = ctx.pipeline
+
+        ori_md_info_list: List[Dict[str, Any]] = []
+        for i, item in enumerate(request.markdownList):
+            ori_md_info_list.append(
+                {
+                    "input_path": None,
+                    "page_index": i,
+                    "markdown_texts": item.text,
+                    "page_continuation_flags": (item.isStart, item.isEnd),
+                }
+            )
+
+        result = await serving_utils.call_async(
+            pipeline.pipeline.translate,
+            ori_md_info_list,
+            target_language=request.targetLanguage,
+            chunk_size=request.chunkSize,
+            task_description=request.taskDescription,
+            output_format=request.outputFormat,
+            rules_str=request.rulesStr,
+            few_shot_demo_text_content=request.fewShotDemoTextContent,
+            few_shot_demo_key_value_list=request.fewShotDemoKeyValueList,
+            chat_bot_config=request.chatBotConfig,
+            llm_request_interval=request.llmRequestInterval,
+        )
+
+        translation_results: List[Dict[str, Any]] = []
+        for item in result:
+            translation_results.append(
+                dict(
+                    language=item["language"],
+                    markdown=dict(
+                        text=item["markdown_texts"],
+                        isStart=item["page_continuation_flags"][0],
+                        isEnd=item["page_continuation_flags"][1],
+                    ),
+                )
+            )
+
+        return AIStudioResultResponse[schema.TranslateResult](
+            logId=serving_utils.generate_log_id(),
+            result=schema.TranslateResult(
+                translationResults=translation_results,
+            ),
+        )
+
+    return app

paddlex/inference/serving/basic_serving/_pipeline_apps/pp_shituv2.py

@@ -161,7 +161,9 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
     ) -> AIStudioResultResponse[schema.InferResult]:
         pipeline = ctx.pipeline
         aiohttp_session = ctx.aiohttp_session
-
+        visualize_enabled = (
+            request.visualize if request.visualize is not None else ctx.config.visualize
+        )
         image_bytes = await serving_utils.get_raw_bytes_async(
             request.image, aiohttp_session
         )
@@ -206,7 +208,7 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
                     score=obj["det_score"],
                 )
             )
-        if ctx.config.visualize:
+        if visualize_enabled:
             output_image_base64 = serving_utils.base64_encode(
                 serving_utils.image_to_bytes(result.img["res"])
             )

paddlex/inference/serving/basic_serving/_pipeline_apps/pp_structurev3.py

@@ -46,7 +46,9 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
         pipeline = ctx.pipeline
 
         log_id = serving_utils.generate_log_id()
-
+        visualize_enabled = (
+            request.visualize if request.visualize is not None else ctx.config.visualize
+        )
         images, data_info = await ocr_common.get_images(request, ctx)
 
         result = await pipeline.infer(
@@ -98,7 +100,7 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
                 max_img_size=ctx.extra["max_output_img_size"],
             )
             md_flags = md_data["page_continuation_flags"]
-            if ctx.config.visualize:
+            if visualize_enabled:
                 imgs = {
                     "input_img": img,
                     **item.img,

paddlex/inference/serving/basic_serving/_pipeline_apps/rotated_object_detection.py

@@ -43,7 +43,9 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
     async def _infer(request: InferRequest) -> AIStudioResultResponse[InferResult]:
         pipeline = ctx.pipeline
         aiohttp_session = ctx.aiohttp_session
-
+        visualize_enabled = (
+            request.visualize if request.visualize is not None else ctx.config.visualize
+        )
         file_bytes = await serving_utils.get_raw_bytes_async(
             request.image, aiohttp_session
         )
@@ -66,7 +68,7 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
                     score=obj["score"],
                 )
             )
-        if ctx.config.visualize:
+        if visualize_enabled:
             output_image_base64 = serving_utils.base64_encode(
                 serving_utils.image_to_bytes(result.img["res"])
             )

paddlex/inference/serving/basic_serving/_pipeline_apps/seal_recognition.py

@@ -44,7 +44,9 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
         pipeline = ctx.pipeline
 
         log_id = serving_utils.generate_log_id()
-
+        visualize_enabled = (
+            request.visualize if request.visualize is not None else ctx.config.visualize
+        )
         images, data_info = await ocr_common.get_images(request, ctx)
 
         result = await pipeline.infer(
@@ -67,7 +69,7 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
         seal_rec_results: List[Dict[str, Any]] = []
         for i, (img, item) in enumerate(zip(images, result)):
             pruned_res = common.prune_result(item.json["res"])
-            if ctx.config.visualize:
+            if visualize_enabled:
                 imgs = {
                     "input_img": img,
                     **item.img,

paddlex/inference/serving/basic_serving/_pipeline_apps/semantic_segmentation.py

@@ -39,7 +39,9 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
     async def _infer(request: InferRequest) -> AIStudioResultResponse[InferResult]:
         pipeline = ctx.pipeline
         aiohttp_session = ctx.aiohttp_session
-
+        visualize_enabled = (
+            request.visualize if request.visualize is not None else ctx.config.visualize
+        )
         file_bytes = await serving_utils.get_raw_bytes_async(
             request.image, aiohttp_session
         )
@@ -50,7 +52,7 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
         pred = result["pred"][0].tolist()
         size = [len(pred), len(pred[0])]
         label_map = [item for sublist in pred for item in sublist]
-        if ctx.config.visualize:
+        if visualize_enabled:
             output_image_base64 = serving_utils.base64_encode(
                 serving_utils.image_to_bytes(result.img["res"].convert("RGB"))
             )

paddlex/inference/serving/basic_serving/_pipeline_apps/small_object_detection.py

@@ -39,7 +39,9 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
     async def _infer(request: InferRequest) -> AIStudioResultResponse[InferResult]:
         pipeline = ctx.pipeline
         aiohttp_session = ctx.aiohttp_session
-
+        visualize_enabled = (
+            request.visualize if request.visualize is not None else ctx.config.visualize
+        )
         file_bytes = await serving_utils.get_raw_bytes_async(
             request.image, aiohttp_session
         )
@@ -57,7 +59,7 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
                     score=obj["score"],
                 )
             )
-        if ctx.config.visualize:
+        if visualize_enabled:
             output_image_base64 = serving_utils.base64_encode(
                 serving_utils.image_to_bytes(result.img["res"])
             )

paddlex/inference/serving/basic_serving/_pipeline_apps/table_recognition.py

@@ -44,7 +44,9 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
         pipeline = ctx.pipeline
 
         log_id = serving_utils.generate_log_id()
-
+        visualize_enabled = (
+            request.visualize if request.visualize is not None else ctx.config.visualize
+        )
         images, data_info = await ocr_common.get_images(request, ctx)
 
         result = await pipeline.infer(
@@ -65,7 +67,7 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
         table_rec_results: List[Dict[str, Any]] = []
         for i, (img, item) in enumerate(zip(images, result)):
             pruned_res = common.prune_result(item.json["res"])
-            if ctx.config.visualize:
+            if visualize_enabled:
                 imgs = {
                     "input_img": img,
                     **item.img,

paddlex/inference/serving/basic_serving/_pipeline_apps/table_recognition_v2.py

@@ -44,7 +44,9 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
         pipeline = ctx.pipeline
 
         log_id = serving_utils.generate_log_id()
-
+        visualize_enabled = (
+            request.visualize if request.visualize is not None else ctx.config.visualize
+        )
         images, data_info = await ocr_common.get_images(request, ctx)
 
         result = await pipeline.infer(
@@ -70,7 +72,7 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
         table_rec_results: List[Dict[str, Any]] = []
         for i, (img, item) in enumerate(zip(images, result)):
             pruned_res = common.prune_result(item.json["res"])
-            if ctx.config.visualize:
+            if visualize_enabled:
                 imgs = {
                     "input_img": img,
                     **item.img,

paddlex/inference/serving/basic_serving/_pipeline_apps/ts_anomaly_detection.py

@@ -39,7 +39,9 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
     async def _infer(request: InferRequest) -> AIStudioResultResponse[InferResult]:
         pipeline = ctx.pipeline
         aiohttp_session = ctx.aiohttp_session
-
+        visualize_enabled = (
+            request.visualize if request.visualize is not None else ctx.config.visualize
+        )
         file_bytes = await serving_utils.get_raw_bytes_async(
             request.csv, aiohttp_session
         )
@@ -50,7 +52,7 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
         output_csv = serving_utils.base64_encode(
             serving_utils.data_frame_to_bytes(result["anomaly"])
         )
-        if ctx.config.visualize:
+        if visualize_enabled:
             output_image = serving_utils.base64_encode(
                 serving_utils.image_to_bytes(result.img["res"].convert("RGB"))
             )

paddlex/inference/serving/basic_serving/_pipeline_apps/ts_classification.py

@@ -39,7 +39,9 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
     async def _infer(request: InferRequest) -> AIStudioResultResponse[InferResult]:
         pipeline = ctx.pipeline
         aiohttp_session = ctx.aiohttp_session
-
+        visualize_enabled = (
+            request.visualize if request.visualize is not None else ctx.config.visualize
+        )
         file_bytes = await serving_utils.get_raw_bytes_async(
             request.csv, aiohttp_session
         )
@@ -49,7 +51,7 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
 
         label = str(result["classification"].at[0, "classid"])
         score = float(result["classification"].at[0, "score"])
-        if ctx.config.visualize:
+        if visualize_enabled:
             output_image = serving_utils.base64_encode(
                 serving_utils.image_to_bytes(result.img["res"].convert("RGB"))
             )

paddlex/inference/serving/basic_serving/_pipeline_apps/ts_forecast.py

@@ -39,7 +39,9 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
     async def _infer(request: InferRequest) -> AIStudioResultResponse[InferResult]:
         pipeline = ctx.pipeline
         aiohttp_session = ctx.aiohttp_session
-
+        visualize_enabled = (
+            request.visualize if request.visualize is not None else ctx.config.visualize
+        )
         file_bytes = await serving_utils.get_raw_bytes_async(
             request.csv, aiohttp_session
         )
@@ -50,7 +52,7 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
         output_csv = serving_utils.base64_encode(
             serving_utils.data_frame_to_bytes(result["forecast"])
         )
-        if ctx.config.visualize:
+        if visualize_enabled:
             output_image = serving_utils.base64_encode(
                 serving_utils.image_to_bytes(result.img["res"].convert("RGB"))
             )

paddlex/inference/serving/basic_serving/_pipeline_apps/vehicle_attribute_recognition.py

@@ -43,7 +43,9 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
     async def _infer(request: InferRequest) -> AIStudioResultResponse[InferResult]:
         pipeline = ctx.pipeline
         aiohttp_session = ctx.aiohttp_session
-
+        visualize_enabled = (
+            request.visualize if request.visualize is not None else ctx.config.visualize
+        )
         file_bytes = await serving_utils.get_raw_bytes_async(
             request.image, aiohttp_session
         )
@@ -69,7 +71,7 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
                     score=obj["det_score"],
                 )
             )
-        if ctx.config.visualize:
+        if visualize_enabled:
             output_image_base64 = serving_utils.base64_encode(
                 serving_utils.image_to_bytes(result.img["res"])
             )

paddlex/inference/serving/infra/utils.py

@@ -18,6 +18,7 @@ import io
 import mimetypes
 import re
 import tempfile
+import threading
 import uuid
 from functools import partial
 from typing import Awaitable, Callable, List, Optional, Tuple, TypeVar, Union, overload
@@ -176,29 +177,33 @@ def base64_encode(data: bytes) -> str:
     return base64.b64encode(data).decode("ascii")
 
 
+_lock = threading.Lock()
+
+
 @function_requires_deps("pypdfium2", "opencv-contrib-python")
 def read_pdf(
     bytes_: bytes, max_num_imgs: Optional[int] = None
 ) -> Tuple[List[np.ndarray], PDFInfo]:
     images: List[np.ndarray] = []
     page_info_list: List[PDFPageInfo] = []
-    doc = pdfium.PdfDocument(bytes_)
-    for page in doc:
-        if max_num_imgs is not None and len(images) >= max_num_imgs:
-            break
-        # TODO: Do not always use zoom=2.0
-        zoom = 2.0
-        deg = 0
-        image = page.render(scale=zoom, rotation=deg).to_pil()
-        image = image.convert("RGB")
-        image = np.array(image)
-        image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
-        images.append(image)
-        page_info = PDFPageInfo(
-            width=image.shape[1],
-            height=image.shape[0],
-        )
-        page_info_list.append(page_info)
+    with _lock:
+        doc = pdfium.PdfDocument(bytes_)
+        for page in doc:
+            if max_num_imgs is not None and len(images) >= max_num_imgs:
+                break
+            # TODO: Do not always use zoom=2.0
+            zoom = 2.0
+            deg = 0
+            image = page.render(scale=zoom, rotation=deg).to_pil()
+            image = image.convert("RGB")
+            image = np.array(image)
+            image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
+            images.append(image)
+            page_info = PDFPageInfo(
+                width=image.shape[1],
+                height=image.shape[0],
+            )
+            page_info_list.append(page_info)
     pdf_info = PDFInfo(
         numPages=len(page_info_list),
         pages=page_info_list,

paddlex/inference/serving/schemas/anomaly_detection.py

@@ -26,6 +26,7 @@ INFER_ENDPOINT: Final[str] = "/image-anomaly-detection"
 
 class InferRequest(BaseModel):
     image: str
+    visualize: Optional[bool] = None
 
 
 class InferResult(BaseModel):

paddlex/inference/serving/schemas/doc_preprocessor.py

@@ -35,6 +35,7 @@ class InferRequest(ocr.BaseInferRequest):
     # consistent with the parameters of the wrapped function though.
     useDocOrientationClassify: Optional[bool] = None
     useDocUnwarping: Optional[bool] = None
+    visualize: Optional[bool] = None
 
 
 class DocPreprocessingResult(BaseModel):

paddlex/inference/serving/schemas/face_recognition.py

@@ -90,6 +90,7 @@ class InferRequest(BaseModel):
     recThreshold: Optional[float] = None
     hammingRadius: Optional[float] = None
     topk: Optional[int] = None
+    visualize: Optional[bool] = None
 
 
 class RecResult(BaseModel):

paddlex/inference/serving/schemas/formula_recognition.py

@@ -38,6 +38,7 @@ class InferRequest(ocr.BaseInferRequest):
     layoutNms: Optional[bool] = None
     layoutUnclipRatio: Optional[Union[float, Tuple[float, float]]] = None
     layoutMergeBboxesMode: Optional[str] = None
+    visualize: Optional[bool] = None
 
 
 class FormulaRecResult(BaseModel):

paddlex/inference/serving/schemas/human_keypoint_detection.py

@@ -36,6 +36,7 @@ INFER_ENDPOINT: Final[str] = "/human-keypoint-detection"
 class InferRequest(BaseModel):
     image: str
     detThreshold: Optional[float] = None
+    visualize: Optional[bool] = None
 
 
 class Person(BaseModel):

paddlex/inference/serving/schemas/image_classification.py

@@ -33,6 +33,7 @@ INFER_ENDPOINT: Final[str] = "/image-classification"
 class InferRequest(BaseModel):
     image: str
     topk: Optional[Annotated[int, Field(gt=0)]] = None
+    visualize: Optional[bool] = None
 
 
 class InferResult(BaseModel):

paddlex/inference/serving/schemas/image_multilabel_classification.py

@@ -32,6 +32,7 @@ INFER_ENDPOINT: Final[str] = "/multilabel-image-classification"
 
 class InferRequest(BaseModel):
     image: str
+    visualize: Optional[bool] = None
     threshold: Optional[
         Union[float, Dict[Union[Literal["default"], int], float], List[float]]
     ] = None