feat: make table transformer parameters configurable (#224)

- refactor `tables.py` so that the structure element confidence
threshold values are loaded from `inference_config`
- refactor intersection over box area threshold in
`objects_to_structure` to config intead of using hardwired value of 0.5
(default is still 0.5)

Author: badGarnet

CHANGELOG.md

@@ -1,3 +1,7 @@
+## 0.6.2
+
+* move the confidence threshold for table transformer to config
+
 ## 0.6.1
 
 * YoloX_quantized is now the default model. This models detects most diverse types and detect tables better than previous model.

unstructured_inference/__version__.py

@@ -1 +1 @@
-__version__ = "0.6.1"  # pragma: no cover
+__version__ = "0.6.2"  # pragma: no cover

unstructured_inference/config.py

@@ -46,6 +46,42 @@ def TABLE_IMAGE_BACKGROUND_PAD(self) -> int:
         """
         return self._get_int("TABLE_IMAGE_BACKGROUND_PAD", 0)
 
+    @property
+    def TT_TABLE_CONF(self) -> float:
+        """confidence threshold for table identified by table transformer"""
+        return self._get_float("TT_TABLE_CONF", 0.5)
+
+    @property
+    def TABLE_COLUMN_CONF(self) -> float:
+        """confidence threshold for column identified by table transformer"""
+        return self._get_float("TABLE_COLUMN_CONF", 0.5)
+
+    @property
+    def TABLE_ROW_CONF(self) -> float:
+        """confidence threshold for column identified by table transformer"""
+        return self._get_float("TABLE_ROW_CONF", 0.5)
+
+    @property
+    def TABLE_COLUMN_HEADER_CONF(self) -> float:
+        """confidence threshold for column header identified by table transformer"""
+        return self._get_float("TABLE_COLUMN_HEADER_CONF", 0.5)
+
+    @property
+    def TABLE_PROJECTED_ROW_HEADER_CONF(self) -> float:
+        """confidence threshold for projected row header identified by table transformer"""
+        return self._get_float("TABLE_PROJECTED_ROW_HEADER_CONF", 0.5)
+
+    @property
+    def TABLE_SPANNING_CELL_CONF(self) -> float:
+        """confidence threshold for table spanning cells identified by table transformer"""
+        return self._get_float("TABLE_SPANNING_CELL_CONF", 0.5)
+
+    @property
+    def TABLE_IOB_THRESHOLD(self) -> float:
+        """minimum intersection over box area ratio for a box to be considered part of a larger box
+        it intersects"""
+        return self._get_float("TABLE_IOB_THRESHOLD", 0.5)
+
     @property
     def LAYOUT_SAME_REGION_THRESHOLD(self) -> float:
         """threshold for two layouts' bounding boxes to be considered as the same region

unstructured_inference/models/tables.py

@@ -177,12 +177,13 @@ def get_class_map(data_type: str):
 
 
 structure_class_thresholds = {
-    "table": 0.5,
-    "table column": 0.5,
-    "table row": 0.5,
-    "table column header": 0.5,
-    "table projected row header": 0.5,
-    "table spanning cell": 0.5,
+    "table": inference_config.TT_TABLE_CONF,
+    "table column": inference_config.TABLE_COLUMN_CONF,
+    "table row": inference_config.TABLE_ROW_CONF,
+    "table column header": inference_config.TABLE_COLUMN_HEADER_CONF,
+    "table projected row header": inference_config.TABLE_PROJECTED_ROW_HEADER_CONF,
+    "table spanning cell": inference_config.TABLE_SPANNING_CELL_CONF,
+    # FIXME (yao) this parameter doesn't seem to be used at all in inference? Can we remove it
     "no object": 10,
 }
 
@@ -282,8 +283,16 @@ def objects_to_structures(objects, tokens, class_thresholds):
     table_structures = []
 
     for table in tables:
-        table_objects = [obj for obj in objects if iob(obj["bbox"], table["bbox"]) >= 0.5]
-        table_tokens = [token for token in tokens if iob(token["bbox"], table["bbox"]) >= 0.5]
+        table_objects = [
+            obj
+            for obj in objects
+            if iob(obj["bbox"], table["bbox"]) >= inference_config.TABLE_IOB_THRESHOLD
+        ]
+        table_tokens = [
+            token
+            for token in tokens
+            if iob(token["bbox"], table["bbox"]) >= inference_config.TABLE_IOB_THRESHOLD
+        ]
 
         structure = {}
 
@@ -302,7 +311,7 @@ def objects_to_structures(objects, tokens, class_thresholds):
         for obj in rows:
             obj["column header"] = False
             for header_obj in column_headers:
-                if iob(obj["bbox"], header_obj["bbox"]) >= 0.5:
+                if iob(obj["bbox"], header_obj["bbox"]) >= inference_config.TABLE_IOB_THRESHOLD:
                     obj["column header"] = True
 
         # Refine table structures
@@ -478,7 +487,7 @@ def structure_to_cells(table_structure, tokens):
                 spanning_cell_rect = Rect(list(spanning_cell["bbox"]))
                 if (
                     spanning_cell_rect.intersect(cell_rect).get_area() / cell_rect.get_area()
-                ) > 0.5:
+                ) > inference_config.TABLE_IOB_THRESHOLD:
                     cell["subcell"] = True
                     break
 
@@ -499,7 +508,9 @@ def structure_to_cells(table_structure, tokens):
         for subcell in subcells:
             subcell_rect = Rect(list(subcell["bbox"]))
             subcell_rect_area = subcell_rect.get_area()
-            if (subcell_rect.intersect(spanning_cell_rect).get_area() / subcell_rect_area) > 0.5:
+            if (
+                subcell_rect.intersect(spanning_cell_rect).get_area() / subcell_rect_area
+            ) > inference_config.TABLE_IOB_THRESHOLD:
                 if cell_rect is None:
                     cell_rect = Rect(list(subcell["bbox"]))
                 else: