Adding EffOCR support within the layout parser

src/layoutparser/__init__.py → src/effocr-layout/__init__.py

@@ -23,6 +23,7 @@
     is_effdet_available,
     is_pytesseract_available,
     is_gcv_available,
+    is_effocr_available,
 )
 
 _import_structure = {
@@ -51,6 +52,7 @@
         "is_paddle_available",
         "is_pytesseract_available",
         "is_gcv_available",
+        "is_effocr_available",
         "requires_backends"
     ],
     "tools": [
@@ -80,6 +82,9 @@
 if is_gcv_available():
     _import_structure["ocr.gcv_agent"] = ["GCVAgent", "GCVFeatureType"]
 
+if is_effocr_available():
+    _import_structure["ocr.effocr_agent"] = ["EffOCRAgent", "EffOCRFeatureType"]
+
 sys.modules[__name__] = _LazyModule(
     __name__,
     globals()["__file__"],

src/layoutparser/file_utils.py → src/effocr-layout/file_utils.py

@@ -88,6 +88,14 @@
 except ModuleNotFoundError:
     _gcv_available = False
 
+try:
+    _effocr_available = importlib.util.find_spec("onnxruntime") is not None \
+                    and importlib.util.find_spec("onnx") is not None \
+                    and importlib.util.find_spec("faiss") is not None  
+except ModuleNotFoundError:
+    _effocr_available = False
+
+
 
 def is_torch_available():
     return _torch_available
@@ -121,6 +129,9 @@ def is_pytesseract_available():
 def is_gcv_available():
     return _gcv_available
 
+def is_effocr_available():
+    return _effocr_available
+
 
 PYTORCH_IMPORT_ERROR = """
 {0} requires the PyTorch library but it was not found in your environment. Checkout the instructions on the
@@ -154,6 +165,13 @@ def is_gcv_available():
 `pip install google-cloud-vision==1`
 """
 
+EFFOCR_IMPORT_ERROR = """
+{0} requires the onnxruntime, onnx and faiss libraries but at least one was not found in your environment. You can install it with pip:
+`pip install onnxruntime onnx faiss`
+Note that `faiss` can be installed with eiter the CPU or GPU version, but the GPU version requires CUDA. See 
+https://github.com/facebookresearch/faiss/blob/main/INSTALL.md for more details.
+"""
+
 BACKENDS_MAPPING = dict(
     [
         ("torch", (is_torch_available, PYTORCH_IMPORT_ERROR)),
@@ -162,6 +180,7 @@ def is_gcv_available():
         ("effdet", (is_effdet_available, EFFDET_IMPORT_ERROR)),
         ("pytesseract", (is_pytesseract_available, PYTESSERACT_IMPORT_ERROR)),
         ("google-cloud-vision", (is_gcv_available, GCV_IMPORT_ERROR)),
+        ("effocr", (is_effocr_available, EFFOCR_IMPORT_ERROR))
     ]
 )
 

src/layoutparser/ocr/__init__.py → src/effocr-layout/ocr/__init__.py

@@ -13,4 +13,5 @@
 # limitations under the License.
 
 from .gcv_agent import GCVAgent, GCVFeatureType
-from .tesseract_agent import TesseractAgent, TesseractFeatureType
+from .tesseract_agent import TesseractAgent, TesseractFeatureType
+from .effocr_agent import EffOCRAgent, EffOCRFeatureType

src/effocr-layout/ocr/effocr/__init__.py

@@ -0,0 +1,3 @@
+from .engines import EffLineDetector, EffLocalizer, EffRecognizer
+from .utils import create_paired_transform, create_paired_transform_word, letterbox, non_max_suppression
+from .infer_transcripton import run_effocr_word

src/effocr-layout/ocr/effocr/engines/__init__.py

@@ -0,0 +1,3 @@
+from .localizer_engine import EffLocalizer
+from .recognizer_engine import EffRecognizer
+from .line_det_engine import EffLineDetector