enhancement: warn if low dpi chipper (#171)

Adds a warning if Chipper is used with DPI less than 300.

Author: qued

CHANGELOG.md

@@ -1,3 +1,7 @@
+## 0.5.11-dev0
+
+* Add warning when chipper is used with < 300 DPI
+
 ## 0.5.10
 
 * Implement full-page OCR

test_unstructured_inference/inference/test_layout.py

@@ -11,7 +11,7 @@
 import unstructured_inference.models.base as models
 from unstructured_inference.inference import elements, layout, layoutelement
 from unstructured_inference.inference.layout import create_image_output_dir
-from unstructured_inference.models import detectron2, tesseract
+from unstructured_inference.models import chipper, detectron2, tesseract
 from unstructured_inference.models.unstructuredmodel import (
     UnstructuredElementExtractionModel,
     UnstructuredObjectDetectionModel,
@@ -866,3 +866,14 @@ def test_create_image_output_dir_no_ext():
         assert os.path.isdir(output_dir)
         assert os.path.isabs(output_dir)
         assert output_dir == expected_output_dir
+
+
+def test_warning_if_chipper_and_low_dpi(caplog):
+    with patch.object(layout.DocumentLayout, "from_file") as mock_from_file, patch.object(
+        chipper.UnstructuredChipperModel,
+        "initialize",
+    ):
+        layout.process_file_with_model("asdf", model_name="chipper", pdf_image_dpi=299)
+        mock_from_file.assert_called_once()
+        assert caplog.records[0].levelname == "WARNING"
+        assert "DPI >= 300" in caplog.records[0].msg

test_unstructured_inference/models/test_model.py

@@ -42,8 +42,9 @@ def test_get_model_warns_on_chipper(monkeypatch, caplog):
         "UnstructuredChipperModel",
         MockModel,
     )
-    models.get_model("chipper")
-    assert caplog.records[0].levelname == "WARNING"
+    with mock.patch.object(models, "models", {}):
+        models.get_model("chipper")
+        assert caplog.records[0].levelname == "WARNING"
 
 
 def test_raises_invalid_model():

unstructured_inference/__version__.py

@@ -1 +1 @@
-__version__ = "0.5.10"  # pragma: no cover
+__version__ = "0.5.11-dev0"  # pragma: no cover

unstructured_inference/inference/layout.py

@@ -401,6 +401,13 @@ def process_file_with_model(
 ) -> DocumentLayout:
     """Processes pdf file with name filename into a DocumentLayout by using a model identified by
     model_name."""
+
+    if (pdf_image_dpi < 300) and (model_name == "chipper"):
+        logger.warning(
+            "The Chipper model performs better when images are rendered with DPI >= 300 "
+            f"(currently {pdf_image_dpi}).",
+        )
+
     model = get_model(model_name)
     if isinstance(model, UnstructuredObjectDetectionModel):
         detection_model = model