Support setting PDF rendering scale factor (#4967)

paddlex/inference/common/batch_sampler/image_batch_sampler.py

@@ -20,6 +20,7 @@
 from ....utils import logging
 from ....utils.cache import CACHE_DIR
 from ....utils.download import download
+from ....utils.flags import PDF_RENDER_SCALE
 from ...utils.io import PDFReader
 from .base_batch_sampler import BaseBatchSampler, Batch
 
@@ -48,7 +49,7 @@ class ImageBatchSampler(BaseBatchSampler):
 
     def __init__(self, *args, **kwargs):
         super().__init__(*args, **kwargs)
-        self.pdf_reader = PDFReader()
+        self.pdf_reader = PDFReader(zoom=PDF_RENDER_SCALE)
 
     # XXX: auto download for url
     def _download_from_url(self, in_path):

paddlex/inference/serving/infra/utils.py

@@ -30,7 +30,7 @@
 from typing_extensions import Literal, ParamSpec, TypeAlias, assert_never
 
 from ....utils.deps import function_requires_deps, is_dep_available
-from ...utils.pdfium_lock import pdfium_lock
+from ...utils.pdfium_lock import PDF_RENDER_SCALE, pdfium_lock
 from .models import ImageInfo, PDFInfo, PDFPageInfo
 
 if is_dep_available("aiohttp"):
@@ -191,8 +191,7 @@ def read_pdf(
                 if max_num_imgs is not None and len(images) >= max_num_imgs:
                     page.close()
                     break
-                # TODO: Do not always use zoom=2.0
-                zoom = 2.0
+                zoom = PDF_RENDER_SCALE
                 deg = 0
                 image = page.render(scale=zoom, rotation=deg).to_numpy()
                 images.append(image)

paddlex/utils/flags.py

@@ -70,7 +70,9 @@ def get_flag_from_env_var(name, default, format_func=str):
     "PADDLE_PDX_DISABLE_MODEL_SOURCE_CHECK", False
 )
 
-HUGGING_FACE_ENDPOINT = os.environ.get("PADDLE_PDX_HUGGING_FACE_ENDPOINT", "https://huggingface.co")
+HUGGING_FACE_ENDPOINT = os.environ.get(
+    "PADDLE_PDX_HUGGING_FACE_ENDPOINT", "https://huggingface.co"
+)
 
 # Inference Benchmark
 INFER_BENCHMARK = get_flag_from_env_var("PADDLE_PDX_INFER_BENCHMARK", False)
@@ -87,3 +89,5 @@ def get_flag_from_env_var(name, default, format_func=str):
 INFER_BENCHMARK_USE_CACHE_FOR_READ = get_flag_from_env_var(
     "PADDLE_PDX_INFER_BENCHMARK_USE_CACHE_FOR_READ", False
 )
+
+PDF_RENDER_SCALE = get_flag_from_env_var("PADDLE_PDX_PDF_RENDER_SCALE", 2.0, float)