fix[tableformer]: Fix cell bbox coordinates (#2)

Signed-off-by: Nikos Livathinos <[email protected]>

Author: nikos-livathinos

docling_ibm_models/tableformer/data_management/tf_predictor.py

@@ -437,10 +437,10 @@ def _merge_tf_output(self, docling_output, pdf_cells):
                 for pdf_cell in pdf_cells:
                     if pdf_cell["id"] == docling_item["cell_id"]:
                         text_cell_bbox = {
-                            "b": pdf_cell["bbox"][1],
+                            "b": pdf_cell["bbox"][3],
                             "l": pdf_cell["bbox"][0],
                             "r": pdf_cell["bbox"][2],
-                            "t": pdf_cell["bbox"][3],
+                            "t": pdf_cell["bbox"][1],
                             "token": pdf_cell["text"],
                         }
                         tf_cells_map[cell_key]["text_cell_bboxes"].append(
@@ -468,10 +468,10 @@ def _merge_tf_output(self, docling_output, pdf_cells):
                 for pdf_cell in pdf_cells:
                     if pdf_cell["id"] == docling_item["cell_id"]:
                         text_cell_bbox = {
-                            "b": pdf_cell["bbox"][1],
+                            "b": pdf_cell["bbox"][3],
                             "l": pdf_cell["bbox"][0],
                             "r": pdf_cell["bbox"][2],
-                            "t": pdf_cell["bbox"][3],
+                            "t": pdf_cell["bbox"][1],
                             "token": pdf_cell["text"],
                         }
                         tf_cells_map[cell_key]["text_cell_bboxes"].append(

tests/test_tf_predictor.py

@@ -564,9 +564,9 @@ def test_tf_predictor():
                         # Visualization:
                         for text_cell in response["text_cell_bboxes"]:
                             xc0 = text_cell["l"]
-                            yc0 = text_cell["b"]
+                            yc0 = text_cell["t"]
                             xc1 = text_cell["r"]
-                            yc1 = text_cell["t"]
+                            yc1 = text_cell["b"]
                             img1.rectangle(((xc0, yc0), (xc1, yc1)), outline="red")
 
                         x0 = response["bbox"]["l"] - 6