onnxtr 0.1.2__py3-none-any.whl → 0.3.0__py3-none-any.whl

onnxtr/models/recognition/models/vitstr.py

@@ -11,7 +11,7 @@ from scipy.special import softmax
 
 from onnxtr.utils import VOCABS
 
-from ...engine import Engine
+from ...engine import Engine, EngineConfig
 from ..core import RecognitionPostProcessor
 
 __all__ = ["ViTSTR", "vitstr_small", "vitstr_base"]
@@ -23,6 +23,7 @@ default_cfgs: Dict[str, Dict[str, Any]] = {
         "input_shape": (3, 32, 128),
         "vocab": VOCABS["french"],
         "url": "https://github.com/felixdittrich92/OnnxTR/releases/download/v0.0.1/vitstr_small-3ff9c500.onnx",
+        "url_8_bit": "https://github.com/felixdittrich92/OnnxTR/releases/download/v0.1.2/vitstr_small_dynamic_8_bit-bec6c796.onnx",
     },
     "vitstr_base": {
         "mean": (0.694, 0.695, 0.693),
@@ -30,6 +31,7 @@ default_cfgs: Dict[str, Dict[str, Any]] = {
         "input_shape": (3, 32, 128),
         "vocab": VOCABS["french"],
         "url": "https://github.com/felixdittrich92/OnnxTR/releases/download/v0.0.1/vitstr_base-ff62f5be.onnx",
+        "url_8_bit": "https://github.com/felixdittrich92/OnnxTR/releases/download/v0.1.2/vitstr_base_dynamic_8_bit-976c7cd6.onnx",
     },
 }
 
@@ -41,6 +43,7 @@ class ViTSTR(Engine):
     ----
         model_path: path to onnx model file
         vocab: vocabulary used for encoding
+        engine_cfg: configuration for the inference engine
         cfg: dictionary containing information about the model
         **kwargs: additional arguments to be passed to `Engine`
     """
@@ -49,10 +52,11 @@ class ViTSTR(Engine):
         self,
         model_path: str,
         vocab: str,
+        engine_cfg: EngineConfig = EngineConfig(),
         cfg: Optional[Dict[str, Any]] = None,
         **kwargs: Any,
     ) -> None:
-        super().__init__(url=model_path, **kwargs)
+        super().__init__(url=model_path, engine_cfg=engine_cfg, **kwargs)
         self.vocab = vocab
         self.cfg = cfg
 
@@ -109,6 +113,8 @@ class ViTSTRPostProcessor(RecognitionPostProcessor):
 def _vitstr(
     arch: str,
     model_path: str,
+    load_in_8_bit: bool = False,
+    engine_cfg: EngineConfig = EngineConfig(),
     **kwargs: Any,
 ) -> ViTSTR:
     # Patch the config
@@ -117,12 +123,19 @@ def _vitstr(
     _cfg["input_shape"] = kwargs.get("input_shape", _cfg["input_shape"])
 
     kwargs["vocab"] = _cfg["vocab"]
+    # Patch the url
+    model_path = default_cfgs[arch]["url_8_bit"] if load_in_8_bit and "http" in model_path else model_path
 
     # Build the model
-    return ViTSTR(model_path, cfg=_cfg, **kwargs)
+    return ViTSTR(model_path, cfg=_cfg, engine_cfg=engine_cfg, **kwargs)
 
 
-def vitstr_small(model_path: str = default_cfgs["vitstr_small"]["url"], **kwargs: Any) -> ViTSTR:
+def vitstr_small(
+    model_path: str = default_cfgs["vitstr_small"]["url"],
+    load_in_8_bit: bool = False,
+    engine_cfg: EngineConfig = EngineConfig(),
+    **kwargs: Any,
+) -> ViTSTR:
     """ViTSTR-Small as described in `"Vision Transformer for Fast and Efficient Scene Text Recognition"
     <https://arxiv.org/pdf/2105.08582.pdf>`_.
 
@@ -135,16 +148,23 @@ def vitstr_small(model_path: str = default_cfgs["vitstr_small"]["url"], **kwargs
     Args:
     ----
         model_path: path to onnx model file, defaults to url in default_cfgs
-        kwargs: keyword arguments of the ViTSTR architecture
+        load_in_8_bit: whether to load the the 8-bit quantized model, defaults to False
+        engine_cfg: configuration for the inference engine
+        **kwargs: keyword arguments of the ViTSTR architecture
 
     Returns:
     -------
         text recognition architecture
     """
-    return _vitstr("vitstr_small", model_path, **kwargs)
+    return _vitstr("vitstr_small", model_path, load_in_8_bit, engine_cfg, **kwargs)
 
 
-def vitstr_base(model_path: str = default_cfgs["vitstr_base"]["url"], **kwargs: Any) -> ViTSTR:
+def vitstr_base(
+    model_path: str = default_cfgs["vitstr_base"]["url"],
+    load_in_8_bit: bool = False,
+    engine_cfg: EngineConfig = EngineConfig(),
+    **kwargs: Any,
+) -> ViTSTR:
     """ViTSTR-Base as described in `"Vision Transformer for Fast and Efficient Scene Text Recognition"
     <https://arxiv.org/pdf/2105.08582.pdf>`_.
 
@@ -157,10 +177,12 @@ def vitstr_base(model_path: str = default_cfgs["vitstr_base"]["url"], **kwargs:
     Args:
     ----
         model_path: path to onnx model file, defaults to url in default_cfgs
-        kwargs: keyword arguments of the ViTSTR architecture
+        load_in_8_bit: whether to load the the 8-bit quantized model, defaults to False
+        engine_cfg: configuration for the inference engine
+        **kwargs: keyword arguments of the ViTSTR architecture
 
     Returns:
     -------
         text recognition architecture
     """
-    return _vitstr("vitstr_base", model_path, **kwargs)
+    return _vitstr("vitstr_base", model_path, load_in_8_bit, engine_cfg, **kwargs)

onnxtr/models/recognition/zoo.py

@@ -5,9 +5,9 @@
 
 from typing import Any, List
 
-from onnxtr.models.preprocessor import PreProcessor
-
 from .. import recognition
+from ..engine import EngineConfig
+from ..preprocessor import PreProcessor
 from .predictor import RecognitionPredictor
 
 __all__ = ["recognition_predictor"]
@@ -25,12 +25,14 @@ ARCHS: List[str] = [
 ]
 
 
-def _predictor(arch: Any, **kwargs: Any) -> RecognitionPredictor:
+def _predictor(
+    arch: Any, load_in_8_bit: bool = False, engine_cfg: EngineConfig = EngineConfig(), **kwargs: Any
+) -> RecognitionPredictor:
     if isinstance(arch, str):
         if arch not in ARCHS:
             raise ValueError(f"unknown architecture '{arch}'")
 
-        _model = recognition.__dict__[arch]()
+        _model = recognition.__dict__[arch](load_in_8_bit=load_in_8_bit, engine_cfg=engine_cfg)
     else:
         if not isinstance(
             arch, (recognition.CRNN, recognition.SAR, recognition.MASTER, recognition.ViTSTR, recognition.PARSeq)
@@ -47,7 +49,9 @@ def _predictor(arch: Any, **kwargs: Any) -> RecognitionPredictor:
     return predictor
 
 
-def recognition_predictor(arch: Any = "crnn_vgg16_bn", **kwargs: Any) -> RecognitionPredictor:
+def recognition_predictor(
+    arch: Any = "crnn_vgg16_bn", load_in_8_bit: bool = False, engine_cfg: EngineConfig = EngineConfig(), **kwargs: Any
+) -> RecognitionPredictor:
     """Text recognition architecture.
 
     Example::
@@ -60,10 +64,12 @@ def recognition_predictor(arch: Any = "crnn_vgg16_bn", **kwargs: Any) -> Recogni
     Args:
     ----
         arch: name of the architecture or model itself to use (e.g. 'crnn_vgg16_bn')
+        load_in_8_bit: whether to load the the 8-bit quantized model, defaults to False
+        engine_cfg: configuration of inference engine
         **kwargs: optional parameters to be passed to the architecture
 
     Returns:
     -------
         Recognition predictor
     """
-    return _predictor(arch, **kwargs)
+    return _predictor(arch, load_in_8_bit, engine_cfg, **kwargs)

onnxtr/models/zoo.py

@@ -6,6 +6,7 @@
 from typing import Any
 
 from .detection.zoo import detection_predictor
+from .engine import EngineConfig
 from .predictor import OCRPredictor
 from .recognition.zoo import recognition_predictor
 
@@ -23,6 +24,10 @@ def _predictor(
     detect_orientation: bool = False,
     straighten_pages: bool = False,
     detect_language: bool = False,
+    load_in_8_bit: bool = False,
+    det_engine_cfg: EngineConfig = EngineConfig(),
+    reco_engine_cfg: EngineConfig = EngineConfig(),
+    clf_engine_cfg: EngineConfig = EngineConfig(),
     **kwargs,
 ) -> OCRPredictor:
     # Detection
@@ -32,12 +37,16 @@ def _predictor(
         assume_straight_pages=assume_straight_pages,
         preserve_aspect_ratio=preserve_aspect_ratio,
         symmetric_pad=symmetric_pad,
+        load_in_8_bit=load_in_8_bit,
+        engine_cfg=det_engine_cfg,
     )
 
     # Recognition
     reco_predictor = recognition_predictor(
         reco_arch,
         batch_size=reco_bs,
+        load_in_8_bit=load_in_8_bit,
+        engine_cfg=reco_engine_cfg,
     )
 
     return OCRPredictor(
@@ -49,6 +58,7 @@ def _predictor(
         detect_orientation=detect_orientation,
         straighten_pages=straighten_pages,
         detect_language=detect_language,
+        clf_engine_cfg=clf_engine_cfg,
         **kwargs,
     )
 
@@ -63,6 +73,10 @@ def ocr_predictor(
     detect_orientation: bool = False,
     straighten_pages: bool = False,
     detect_language: bool = False,
+    load_in_8_bit: bool = False,
+    det_engine_cfg: EngineConfig = EngineConfig(),
+    reco_engine_cfg: EngineConfig = EngineConfig(),
+    clf_engine_cfg: EngineConfig = EngineConfig(),
     **kwargs: Any,
 ) -> OCRPredictor:
     """End-to-end OCR architecture using one model for localization, and another for text recognition.
@@ -94,6 +108,10 @@ def ocr_predictor(
             Doing so will improve performances for documents with page-uniform rotations.
         detect_language: if True, the language prediction will be added to the predictions for each
             page. Doing so will slightly deteriorate the overall latency.
+        load_in_8_bit: whether to load the the 8-bit quantized model, defaults to False
+        det_engine_cfg: configuration of the detection engine
+        reco_engine_cfg: configuration of the recognition engine
+        clf_engine_cfg: configuration of the orientation classification engine
         kwargs: keyword args of `OCRPredictor`
 
     Returns:
@@ -110,5 +128,9 @@ def ocr_predictor(
         detect_orientation=detect_orientation,
         straighten_pages=straighten_pages,
         detect_language=detect_language,
+        load_in_8_bit=load_in_8_bit,
+        det_engine_cfg=det_engine_cfg,
+        reco_engine_cfg=reco_engine_cfg,
+        clf_engine_cfg=clf_engine_cfg,
         **kwargs,
     )

onnxtr/utils/geometry.py

@@ -26,6 +26,7 @@ __all__ = [
     "extract_crops",
     "extract_rcrops",
     "shape_translate",
+    "detach_scores",
 ]
 
 
@@ -58,6 +59,26 @@ def polygon_to_bbox(polygon: Polygon4P) -> BoundingBox:
     return (min(x), min(y)), (max(x), max(y))
 
 
+def detach_scores(boxes: List[np.ndarray]) -> Tuple[List[np.ndarray], List[np.ndarray]]:
+    """Detach the objectness scores from box predictions
+    Args:
+    ----
+        boxes: list of arrays with boxes of shape (N, 5) or (N, 5, 2)
+    Returns:
+    -------
+        a tuple of two lists: the first one contains the boxes without the objectness scores,
+        the second one contains the objectness scores
+    """
+
+    def _detach(boxes: np.ndarray) -> Tuple[np.ndarray, np.ndarray]:
+        if boxes.ndim == 2:
+            return boxes[:, :-1], boxes[:, -1]
+        return boxes[:, :-1], boxes[:, -1, -1]
+
+    loc_preds, obj_scores = zip(*(_detach(box) for box in boxes))
+    return list(loc_preds), list(obj_scores)
+
+
 def shape_translate(data: np.ndarray, format: str) -> np.ndarray:
     """Translate the shape of the input data to the desired format
 
@@ -106,18 +127,18 @@ def resolve_enclosing_bbox(bboxes: Union[List[BoundingBox], np.ndarray]) -> Unio
     ----
         bboxes: boxes in one of the following formats:
 
-            - an array of boxes: (*, 5), where boxes have this shape:
-            (xmin, ymin, xmax, ymax, score)
+            - an array of boxes: (*, 4), where boxes have this shape:
+            (xmin, ymin, xmax, ymax)
 
             - a list of BoundingBox
 
     Returns:
     -------
-        a (1, 5) array (enclosing boxarray), or a BoundingBox
+        a (1, 4) array (enclosing boxarray), or a BoundingBox
     """
     if isinstance(bboxes, np.ndarray):
-        xmin, ymin, xmax, ymax, score = np.split(bboxes, 5, axis=1)
-        return np.array([xmin.min(), ymin.min(), xmax.max(), ymax.max(), score.mean()])
+        xmin, ymin, xmax, ymax = np.split(bboxes, 4, axis=1)
+        return np.array([xmin.min(), ymin.min(), xmax.max(), ymax.max()])
     else:
         x, y = zip(*[point for box in bboxes for point in box])
         return (min(x), min(y)), (max(x), max(y))
@@ -130,21 +151,21 @@ def resolve_enclosing_rbbox(rbboxes: List[np.ndarray], intermed_size: int = 1024
     ----
         rbboxes: boxes in one of the following formats:
 
-            - an array of boxes: (*, 5), where boxes have this shape:
-            (xmin, ymin, xmax, ymax, score)
+            - an array of boxes: (*, 4, 2), where boxes have this shape:
+            (x1, y1), (x2, y2), (x3, y3), (x4, y4)
 
             - a list of BoundingBox
         intermed_size: size of the intermediate image
 
     Returns:
     -------
-        a (1, 5) array (enclosing boxarray), or a BoundingBox
+        a (4, 2) array (enclosing rotated box)
     """
     cloud: np.ndarray = np.concatenate(rbboxes, axis=0)
     # Convert to absolute for minAreaRect
     cloud *= intermed_size
     rect = cv2.minAreaRect(cloud.astype(np.int32))
-    return cv2.boxPoints(rect) / intermed_size  # type: ignore[operator]
+    return cv2.boxPoints(rect) / intermed_size  # type: ignore[return-value]
 
 
 def rotate_abs_points(points: np.ndarray, angle: float = 0.0) -> np.ndarray:
@@ -274,7 +295,7 @@ def rotate_boxes(
 
     Args:
     ----
-        loc_preds: (N, 5) or (N, 4, 2) array of RELATIVE boxes
+        loc_preds: (N, 4) or (N, 4, 2) array of RELATIVE boxes
         angle: angle between -90 and +90 degrees
         orig_shape: shape of the origin image
         min_angle: minimum angle to rotate boxes
@@ -362,7 +383,7 @@ def rotate_image(
             # Pad height
             else:
                 h_pad, w_pad = int(rot_img.shape[1] * image.shape[0] / image.shape[1] - rot_img.shape[0]), 0
-            rot_img = np.pad(rot_img, ((h_pad // 2, h_pad - h_pad // 2), (w_pad // 2, w_pad - w_pad // 2), (0, 0)))
+            rot_img = np.pad(rot_img, ((h_pad // 2, h_pad - h_pad // 2), (w_pad // 2, w_pad - w_pad // 2), (0, 0)))  # type: ignore[assignment]
         if preserve_origin_shape:
             # rescale
             rot_img = cv2.resize(rot_img, image.shape[:-1][::-1], interpolation=cv2.INTER_LINEAR)
@@ -495,4 +516,4 @@ def extract_rcrops(
         )
         for idx in range(_boxes.shape[0])
     ]
-    return crops
+    return crops  # type: ignore[return-value]

onnxtr/version.py

@@ -1 +1 @@
-__version__ = 'v0.1.2'
+__version__ = 'v0.3.0'

{onnxtr-0.1.2.dist-info → onnxtr-0.3.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: onnxtr
-Version: 0.1.2
+Version: 0.3.0
 Summary: Onnx Text Recognition (OnnxTR): docTR Onnx-Wrapper for high-performance OCR on documents.
 Author-email: Felix Dittrich <felixdittrich92@gmail.com>
 Maintainer: Felix Dittrich
@@ -228,7 +228,7 @@ License-File: LICENSE
 Requires-Dist: numpy <2.0.0,>=1.16.0
 Requires-Dist: scipy <2.0.0,>=1.4.0
 Requires-Dist: opencv-python <5.0.0,>=4.5.0
-Requires-Dist: pypdfium2 <5.0.0,>=4.0.0
+Requires-Dist: pypdfium2 <5.0.0,>=4.11.0
 Requires-Dist: pyclipper <2.0.0,>=1.2.0
 Requires-Dist: shapely <3.0.0,>=1.6.0
 Requires-Dist: rapidfuzz <4.0.0,>=3.0.0
@@ -275,7 +275,7 @@ Requires-Dist: mplcursors >=0.3 ; extra == 'viz'
 [![codecov](https://codecov.io/gh/felixdittrich92/OnnxTR/graph/badge.svg?token=WVFRCQBOLI)](https://codecov.io/gh/felixdittrich92/OnnxTR)
 [![Codacy Badge](https://app.codacy.com/project/badge/Grade/4fff4d764bb14fb8b4f4afeb9587231b)](https://app.codacy.com/gh/felixdittrich92/OnnxTR/dashboard?utm_source=gh&utm_medium=referral&utm_content=&utm_campaign=Badge_grade)
 [![CodeFactor](https://www.codefactor.io/repository/github/felixdittrich92/onnxtr/badge)](https://www.codefactor.io/repository/github/felixdittrich92/onnxtr)
-[![Pypi](https://img.shields.io/badge/pypi-v0.1.1-blue.svg)](https://pypi.org/project/OnnxTR/)
+[![Pypi](https://img.shields.io/badge/pypi-v0.3.0-blue.svg)](https://pypi.org/project/OnnxTR/)
 
 > :warning: Please note that this is a wrapper around the [doctr](https://github.com/mindee/doctr) library to provide a Onnx pipeline for docTR. For feature requests, which are not directly related to the Onnx pipeline, please refer to the base project.
 
@@ -284,8 +284,9 @@ Requires-Dist: mplcursors >=0.3 ; extra == 'viz'
 What you can expect from this repository:
 
 - efficient ways to parse textual information (localize and identify each word) from your documents
-- a Onnx pipeline for docTR, a wrapper around the [doctr](https://github.com/mindee/doctr) library
+- a Onnx pipeline for docTR, a wrapper around the [doctr](https://github.com/mindee/doctr) library - no PyTorch or TensorFlow dependencies
 - more lightweight package with faster inference latency and less required resources
+- 8-Bit quantized models for faster inference on CPU
 
 ![OCR_example](https://github.com/felixdittrich92/OnnxTR/raw/main/docs/images/ocr.png)
 
@@ -335,11 +336,11 @@ multi_img_doc = DocumentFile.from_images(["path/to/page1.jpg", "path/to/page2.jp
 
 ### Putting it together
 
-Let's use the default pretrained model for an example:
+Let's use the default `ocr_predictor` model for an example:
 
 ```python
 from onnxtr.io import DocumentFile
-from onnxtr.models import ocr_predictor
+from onnxtr.models import ocr_predictor, EngineConfig
 
 model = ocr_predictor(
     det_arch='fast_base',  # detection architecture
@@ -356,8 +357,15 @@ model = ocr_predictor(
     detect_language=False, # set to `True` if the language of the pages should be detected (default: False)
     # DocumentBuilder specific parameters
     resolve_lines=True,  # whether words should be automatically grouped into lines (default: True)
-    resolve_blocks=True,  # whether lines should be automatically grouped into blocks (default: True)
+    resolve_blocks=False,  # whether lines should be automatically grouped into blocks (default: False)
     paragraph_break=0.035,  # relative length of the minimum space separating paragraphs (default: 0.035)
+    # OnnxTR specific parameters
+    # NOTE: 8-Bit quantized models are not available for FAST detection models and can in general lead to poorer accuracy
+    load_in_8_bit=False,  # set to `True` to load 8-bit quantized models instead of the full precision onces (default: False)
+    # Advanced engine configuration options
+    det_engine_cfg=EngineConfig(),  # detection model engine configuration (default: internal predefined configuration)
+    reco_engine_cfg=EngineConfig(),  # recognition model engine configuration (default: internal predefined configuration)
+    clf_engine_cfg=EngineConfig(),  # classification (orientation) model engine configuration (default: internal predefined configuration)
 )
 # PDF
 doc = DocumentFile.from_pdf("path/to/your/doc.pdf")
@@ -395,6 +403,39 @@ for output in xml_output:
 
 ```
 
+<details>
+  <summary>Advanced engine configuration options</summary>
+
+You can also define advanced engine configurations for the models / predictors:
+
+```python
+from onnxruntime import SessionOptions
+
+from onnxtr.models import ocr_predictor, EngineConfig
+
+general_options = SessionOptions()  # For configuartion options see: https://onnxruntime.ai/docs/api/python/api_summary.html#sessionoptions
+general_options.enable_cpu_mem_arena = False
+
+# NOTE: The following would force to run only on the GPU if no GPU is available it will raise an error
+# List of strings e.g. ["CUDAExecutionProvider", "CPUExecutionProvider"] or a list of tuples with the provider and its options e.g.
+# [("CUDAExecutionProvider", {"device_id": 0}), ("CPUExecutionProvider", {"arena_extend_strategy": "kSameAsRequested"})]
+providers = [("CUDAExecutionProvider", {"device_id": 0})]  # For available providers see: https://onnxruntime.ai/docs/execution-providers/
+
+engine_config = EngineConfig(
+    session_options=general_options,
+    providers=providers
+)
+# We use the default predictor with the custom engine configuration
+# NOTE: You can define differnt engine configurations for detection, recognition and classification depending on your needs
+predictor = ocr_predictor(
+    det_engine_cfg=engine_config,
+    reco_engine_cfg=engine_config,
+    clf_engine_cfg=engine_config
+)
+```
+
+</details>
+
 ## Loading custom exported models
 
 You can also load docTR custom exported models:
@@ -438,9 +479,9 @@ predictor.list_archs()
             'linknet_resnet18',
             'linknet_resnet34',
             'linknet_resnet50',
-            'fast_tiny',
-            'fast_small',
-            'fast_base'
+            'fast_tiny',  # No 8-bit support
+            'fast_small',  # No 8-bit support
+            'fast_base'  # No 8-bit support
         ],
     'recognition archs':
         [
@@ -469,14 +510,38 @@ NOTE:
 
 ### Benchmarks
 
-The benchmarks was measured on a `i7-14700K Intel CPU`.
+The CPU benchmarks was measured on a `i7-14700K Intel CPU`.
+
+The GPU benchmarks was measured on a `RTX 4080 Nvidia GPU`.
+
+Benchmarking performed on the FUNSD dataset and CORD dataset.
+
+docTR / OnnxTR models used for the benchmarks are `fast_base` (full precision) | `db_resnet50` (8-bit variant) for detection and `crnn_vgg16_bn` for recognition.
+
+The smallest combination in OnnxTR (docTR) of `db_mobilenet_v3_large` and `crnn_mobilenet_v3_small` takes as comparison `~0.17s / Page` on the FUNSD dataset and `~0.12s / Page` on the CORD dataset in **full precision**.
+
+- CPU benchmarks:
+
+|Library                          |FUNSD (199 pages)              |CORD  (900 pages)              |
+|---------------------------------|-------------------------------|-------------------------------|
+|docTR (CPU) - v0.8.1             | ~1.29s / Page                 | ~0.60s / Page                 |
+|**OnnxTR (CPU)** - v0.1.2        | ~0.57s / Page                 | **~0.25s / Page**             |
+|**OnnxTR (CPU) 8-bit** - v0.1.2  | **~0.38s / Page**             | **~0.14s / Page**             |
+|EasyOCR (CPU) - v1.7.1           | ~1.96s / Page                 | ~1.75s / Page                 |
+|**PyTesseract (CPU)** - v0.3.10  | **~0.50s / Page**             | ~0.52s / Page                 |
+|Surya (line) (CPU) - v0.4.4      | ~48.76s / Page                | ~35.49s / Page                |
+|PaddleOCR (CPU) - no cls - v2.7.3| ~1.27s / Page                 | ~0.38s / Page                 |
 
-MORE BENCHMARKS COMING SOON
+- GPU benchmarks:
 
-|Dataset                         |docTR (CPU) - v0.8.1           |OnnxTR (CPU) - v0.1.1          |
-|--------------------------------|-------------------------------|-------------------------------|
-|FUNSD (199 pages)               | ~1.29s / Page                 | ~0.57s / Page                 |
-|CORD  (900 pages)               | ~0.60s / Page                 | ~0.25s / Page                 |
+|Library                              |FUNSD (199 pages)              |CORD  (900 pages)              |
+|-------------------------------------|-------------------------------|-------------------------------|
+|docTR (GPU) - v0.8.1                 | ~0.07s / Page                 | ~0.05s / Page                 |
+|**docTR (GPU) float16** - v0.8.1     | **~0.06s / Page**             | **~0.03s / Page**             |
+|OnnxTR (GPU) - v0.1.2                | **~0.06s / Page**             | ~0.04s / Page                 |
+|EasyOCR (GPU) - v1.7.1               | ~0.31s / Page                 | ~0.19s / Page                 |
+|Surya (GPU) float16 - v0.4.4         | ~3.70s / Page                 | ~2.81s / Page                 |
+|**PaddleOCR (GPU) - no cls - v2.7.3**| ~0.08s / Page                 | **~0.03s / Page**             |
 
 ## Citation
 

{onnxtr-0.1.2.dist-info → onnxtr-0.3.0.dist-info}/RECORD

@@ -1,52 +1,55 @@
 onnxtr/__init__.py,sha256=h7Wc2tuHLsaoCk5xNpEFEK-g11A6SJA7nAasA76TQ_Y,100
 onnxtr/file_utils.py,sha256=WjUKalEdR53aoeIY4e-ihy3r7J_C9qFxL40JHGPfutc,1107
-onnxtr/version.py,sha256=t2ZLy6eqIRREzP9MNv-0UaNqpTXYFxlbMzz0JwDxS9I,23
+onnxtr/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+onnxtr/version.py,sha256=3WzdRDDiKxM8JAvNhW3PVopgIZrHCvYuR4insIGe4bU,23
 onnxtr/contrib/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 onnxtr/contrib/artefacts.py,sha256=tdmfhvfXVRYEH7uj4_hqf2cuUGoTieyNK8bXsD3zHwo,5383
 onnxtr/contrib/base.py,sha256=PoCKtOIgj7u4xl-V0eBVh-QmVeTyk_eEggFHQ8R34AI,3445
 onnxtr/io/__init__.py,sha256=kS7tKGFvzxOCWBOun-Y8n9CsziwRKNynjwpZEUUI03M,106
-onnxtr/io/elements.py,sha256=Gm_rNVUhRNMAJFhiw-DejbG6-AXCoCcTyZAoYeTpfW4,16994
+onnxtr/io/elements.py,sha256=h-IxpFqXrvg-fOhpnOqpGFLdG-lR-xYYIxk3chy_MN8,17769
 onnxtr/io/html.py,sha256=Em_7PjZ56SugJ9bjjcWLCMVe5ee6uUMKeZovNxJFAXw,737
 onnxtr/io/image.py,sha256=4tLTh2bGdA0ohh3a6mV6xD0KqNOtIVi5lJ06XSmeyMI,1759
-onnxtr/io/pdf.py,sha256=6ETj8qnUmRLeHP32m0lMJ8EwCaZVEtRoweFGx4Qkrbg,1339
+onnxtr/io/pdf.py,sha256=tD0klmxI-gkMXp56f_ZXWyPHLsUBKa_xlhNTtGV6tpU,1367
 onnxtr/io/reader.py,sha256=BA7DPhW-Gkmce_ZfzrOl4H3pSXVy2JBeQEuY3pWrBFg,2852
-onnxtr/models/__init__.py,sha256=CzdiA34zjFq4a22XvgzG_Bojosi9aN5jeRCizRyh_7o,101
-onnxtr/models/_utils.py,sha256=RnSmj70gFU9CCJlsCYiVULEs7L8MhogkLf2Xu_7wBhM,4824
-onnxtr/models/builder.py,sha256=byUpCs9x5RS2lqNXLECR5GELqd1yF-MEg_09OGWj20Q,13642
-onnxtr/models/engine.py,sha256=2SwzcWCf2s4ti1mHqLF0lqLDw3wpM9FYjdEdtK5MYJ4,2140
-onnxtr/models/zoo.py,sha256=FTG6KCv-4G90N-m0fZhRDOFiyd_sFzLZuBw4bg2ecvI,4302
+onnxtr/models/__init__.py,sha256=Rg-5P2e622q-5ScfxVE3G8GXa51HUPS7b0jkvdukFzM,134
+onnxtr/models/_utils.py,sha256=KncsNcoWqbsxFwduce2STuGHLhv63nXEHv7CMuh6wYA,6606
+onnxtr/models/builder.py,sha256=Bzg-XHZc5k16Ti2XeV9hm4POTHofe581Azq1a3d1O6E,14296
+onnxtr/models/engine.py,sha256=SOK-KTNWMozIjErWQAY56iB2eXyRD44Q08TdL9YOVAY,4717
+onnxtr/models/zoo.py,sha256=MJIT3OZ4kyj2xBfQdCVxl2uBdiLCnnv8czPtHbZl5e4,5343
 onnxtr/models/classification/__init__.py,sha256=h1bZs55iLJBMATtzS4ntTKwfD6OGXBiiqGv_hEnOFnE,41
-onnxtr/models/classification/zoo.py,sha256=yiDTEZCHLBu6WUwQVnQipKhjngPs6OgTeNt4MX8z9m4,2748
+onnxtr/models/classification/zoo.py,sha256=1oaKfW646IVa-MmLqGi58BtBWdHdu4hI8r79wVdLQ2o,3426
 onnxtr/models/classification/models/__init__.py,sha256=rohbM6ZQslfYchi7feZwwh-sX3XXRUhgtEJQeurAytQ,24
-onnxtr/models/classification/models/mobilenet.py,sha256=9SzYYpn1XOUFHZHzKM0BtvSh08WXeAhMeebt3E1eAmQ,3677
+onnxtr/models/classification/models/mobilenet.py,sha256=vTBHhA1okhnCgn36qKlM2eDCm4ftFZDH8Bk2VpkWm4U,4880
 onnxtr/models/classification/predictor/__init__.py,sha256=ERmmOxz_9mUkIuccNbzUa5Y6gVLLVDdyc4cCxbCCUbY,20
-onnxtr/models/classification/predictor/base.py,sha256=9-5i1Uj0EdNQBeSOhaQjGz78Gr0IAr3JZhoKVMA_il4,2009
+onnxtr/models/classification/predictor/base.py,sha256=Xfaj2XlaJuQ2R81XqF5RB0Wcvzd4wh7Z6j1ifn2niFc,2097
 onnxtr/models/detection/__init__.py,sha256=h1bZs55iLJBMATtzS4ntTKwfD6OGXBiiqGv_hEnOFnE,41
 onnxtr/models/detection/core.py,sha256=ZmVDHLJ1l4LQ8rFSKc7enXDkGcOWrcQv4H0SJWyLsag,3584
-onnxtr/models/detection/zoo.py,sha256=GfwvVAT3bbx-1zccqDO-jzOzzL5Sj3D06zbd_x6dUOo,2257
+onnxtr/models/detection/zoo.py,sha256=dpxLC7jMNZyl3a-o4dSCwsMnqtgoRwxy4psZ8WPC6cE,2725
+onnxtr/models/detection/_utils/__init__.py,sha256=oPkIYbySSbLsOk02wVPNO9bUuywC47YjaenfyTwfOsw,20
+onnxtr/models/detection/_utils/base.py,sha256=fOWnvBKluWKTNXSBKg3U6ckzYuF7onEKQ4AvheuTJQk,2346
 onnxtr/models/detection/models/__init__.py,sha256=6Ea6knYrVCR2jAmPlsVWmCdHe-c6lSRETSAuZGfhx8I,85
-onnxtr/models/detection/models/differentiable_binarization.py,sha256=fu1Mn8i38vIyW98YKdPnwXRcvk_-OlkOmcez7wef8l4,5157
-onnxtr/models/detection/models/fast.py,sha256=40P_q5TWUnY9iIyi-pAqnvauQ5KeqIAco2NGGF4plrg,5101
-onnxtr/models/detection/models/linknet.py,sha256=e9uZZB5S3GYImGroHyZ4oonb_ULxL0VAXd_yDuqG1K0,5194
+onnxtr/models/detection/models/differentiable_binarization.py,sha256=o6Y0iDRHxArLqBE-EKz3Ru9l6L7sqHmHkNny60-gV4Q,6734
+onnxtr/models/detection/models/fast.py,sha256=YUnbKLIZdeMd-lfFyWEtRbxpiXsRBizLb0VpcruJD-U,6293
+onnxtr/models/detection/models/linknet.py,sha256=aXOZ6ieczvAoJQcVuVpJZVXqfEIL4OHr5NqQ5nEI2QY,6771
 onnxtr/models/detection/postprocessor/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-onnxtr/models/detection/postprocessor/base.py,sha256=U8FgqveZt2qPHI0aUnIEeX370zGUEWUxL6EPqhFZnRI,5714
+onnxtr/models/detection/postprocessor/base.py,sha256=FIhSNktNLQjGWup3xEMaOCjKQmRvtt0h8M9IFQk_5jM,5823
 onnxtr/models/detection/predictor/__init__.py,sha256=ERmmOxz_9mUkIuccNbzUa5Y6gVLLVDdyc4cCxbCCUbY,20
-onnxtr/models/detection/predictor/base.py,sha256=K0fSTFnA8-fGL9F3HzW3nqnEqO5YKbhFf2TvpUOe_Tw,1744
+onnxtr/models/detection/predictor/base.py,sha256=bt8M6I14tWC9DYjrFrqg-AU5u670_uPpuC7LmcegcCQ,2328
 onnxtr/models/predictor/__init__.py,sha256=XL25XkRkgyK7mldF-CWhg2MMakSdP5vLpDLwL59hphk,25
-onnxtr/models/predictor/base.py,sha256=nrIE6rvxA3LbeUbQ-e3wkochO66OJT3BP-NW_OyHUyw,7555
-onnxtr/models/predictor/predictor.py,sha256=sfwqRzabqavlI6ty_Z6jGm64olmmQBLQGU39_nulQZs,6045
+onnxtr/models/predictor/base.py,sha256=wROvnIvMQb_SPPX8m8_RmBSqZqIDlH7Vfo81D8teQQA,8860
+onnxtr/models/predictor/predictor.py,sha256=kmU6hj89k1QvFpljr3JEWneT7X5RQLcUNn3Ecbb1jm8,6237
 onnxtr/models/preprocessor/__init__.py,sha256=ERmmOxz_9mUkIuccNbzUa5Y6gVLLVDdyc4cCxbCCUbY,20
 onnxtr/models/preprocessor/base.py,sha256=f0t0rMCzvuxwgq7jlKvcVWyjeDOx7yCLUw52quEaETM,3990
 onnxtr/models/recognition/__init__.py,sha256=h1bZs55iLJBMATtzS4ntTKwfD6OGXBiiqGv_hEnOFnE,41
 onnxtr/models/recognition/core.py,sha256=0Q1dVXqRcDUr_ycT5tpoSH9-zuDF58GtnmxWpUS8Ibo,739
 onnxtr/models/recognition/utils.py,sha256=04abbjx-_OuF5iEANWIAOK3tQQl1tExPmBQx4IG04Lc,3569
-onnxtr/models/recognition/zoo.py,sha256=FC8kHSr66s5D-Ltb8dnNI-x3HLdy5LZhZeiWMV69J9k,2115
+onnxtr/models/recognition/zoo.py,sha256=F0hiymT8Tfv115u_34PvmD8rpXw1fPinYno1DE9a8bo,2511
 onnxtr/models/recognition/models/__init__.py,sha256=IXfiuzzkft8O1CpBZWYTpFw19y49mt5rJ_iGSdaWiU0,105
-onnxtr/models/recognition/models/crnn.py,sha256=Y1lmvCAN_r-ciZNHRIZWHLhv60j2zcR1XoRC5r8_BD0,7375
-onnxtr/models/recognition/models/master.py,sha256=4FiWWBtezb7WjhDvnzauZ-iL6_rKashoqbs-9JfHlvM,4009
-onnxtr/models/recognition/models/parseq.py,sha256=qYtXWC8jXnIKViKTLp-WvcEOoU_DPkZ-_KHVhoKgUmw,3765
-onnxtr/models/recognition/models/sar.py,sha256=c0hRZ5dXl9sijkXZXtUnu9cB4DjUFc1MMzriDBhor78,3815
-onnxtr/models/recognition/models/vitstr.py,sha256=jID2RJs5O81b4jBGjkeSNPQ2PTbon-_U8ZDGfCoGcwY,4880
+onnxtr/models/recognition/models/crnn.py,sha256=Ki2DeIQahvIJterFs2RYf-y21LFmFVuhmoem3-nVlXQ,8963
+onnxtr/models/recognition/models/master.py,sha256=VgPwyCpVv6UmTDaeeeGWWgcKPKeEq6Osif-Tq97xmj8,4777
+onnxtr/models/recognition/models/parseq.py,sha256=Ig0Tu31KgVEVWOX630VhEV2hoi5QtABxBrTsgiguK74,4577
+onnxtr/models/recognition/models/sar.py,sha256=OTyXC5_0-DPghHG9zY4ZCnFqAIf-3eBlWoRQOTfjZTc,4588
+onnxtr/models/recognition/models/vitstr.py,sha256=xED7mK1b2d3dUJkLjiFn1JQKe_CU0JE7fhPnEVilT7s,6054
 onnxtr/models/recognition/predictor/__init__.py,sha256=ERmmOxz_9mUkIuccNbzUa5Y6gVLLVDdyc4cCxbCCUbY,20
 onnxtr/models/recognition/predictor/_utils.py,sha256=ZNm5I7ibiWfTlz302uiifCkUOu65YWa-oUBUMPrrUuQ,3406
 onnxtr/models/recognition/predictor/base.py,sha256=YvqSNEM3rCEttxl6hsC9zl1R97N9zO2WZfD5_-nfkR0,2483
@@ -56,15 +59,15 @@ onnxtr/utils/__init__.py,sha256=pESRJKtcQyjRxiMgZPhtPYeLbCj-YSGyMVRHTbcMONU,94
 onnxtr/utils/common_types.py,sha256=eC_NyIwbo9qVF33LiNPqHKfyabWq9mYEKD9gAloo5UU,601
 onnxtr/utils/data.py,sha256=Dh0mgeHJhyPwmm63J90uDVmIYbrp63hh1_SnYLnpgJI,4354
 onnxtr/utils/fonts.py,sha256=OiOHFwkjN4L7QBrzMi7Ex7qj_KcTEJ1sHEJWSfiGNZU,1281
-onnxtr/utils/geometry.py,sha256=PX3e5Cjmb3huwhj2qoMII-8mCxjfWLVzR4O3kSzOdXA,17002
+onnxtr/utils/geometry.py,sha256=u9ei6WW8Yd29rtwnrDYercAY-tWkOLkzBd5Oi6NNyDI,17774
 onnxtr/utils/multithreading.py,sha256=30T7AylM3rb52ZEI3Pk1pfB0VYraTbc7yO2vNODVVFY,2011
 onnxtr/utils/reconstitution.py,sha256=Hx1_ddLevKLzuxXc19UelPdsGlAwqi4f6vRSYKHDUB4,2617
 onnxtr/utils/repr.py,sha256=kfbjGL6KymGT8spo2UL4FJXZ0XRwa7CO7Y1dTVR8dIk,2129
 onnxtr/utils/visualization.py,sha256=CX09qvDnNIw3BFW5F3jM4R9OcpLWAeZyoDyTAOGRvls,9925
 onnxtr/utils/vocabs.py,sha256=SCQ4XQjbHSxunj1tg2iHRiPfE8OaTAMhcJbKq5BNvFs,3138
-onnxtr-0.1.2.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-onnxtr-0.1.2.dist-info/METADATA,sha256=hg64djJ-sUVObh4t9QYB3PJSjZgYxrPne2aS_lEtWcI,25645
-onnxtr-0.1.2.dist-info/WHEEL,sha256=GJ7t_kWBFywbagK5eo9IoUwLW6oyOeTKmQ-9iHFVNxQ,92
-onnxtr-0.1.2.dist-info/top_level.txt,sha256=48aR9EH095hmgye7CNuxAW3o_cj4TjhWmuw3jEMdTs0,12
-onnxtr-0.1.2.dist-info/zip-safe,sha256=AbpHGcgLb-kRsJGnwFEktk7uzpZOCcBY74-YBdrKVGs,1
-onnxtr-0.1.2.dist-info/RECORD,,
+onnxtr-0.3.0.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+onnxtr-0.3.0.dist-info/METADATA,sha256=0cPAKQr-w-WHimev0v9mtys9NetS_oYHZHTslgcSNu4,29756
+onnxtr-0.3.0.dist-info/WHEEL,sha256=mguMlWGMX-VHnMpKOjjQidIo1ssRlCFu4a4mBpz1s2M,91
+onnxtr-0.3.0.dist-info/top_level.txt,sha256=r_MSUTpspp4pWEEWvly-s7ZkfCg1KwrK6-kBlXkWKU8,7
+onnxtr-0.3.0.dist-info/zip-safe,sha256=AbpHGcgLb-kRsJGnwFEktk7uzpZOCcBY74-YBdrKVGs,1
+onnxtr-0.3.0.dist-info/RECORD,,

{onnxtr-0.1.2.dist-info → onnxtr-0.3.0.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: bdist_wheel (0.43.0)
+Generator: setuptools (70.1.1)
 Root-Is-Purelib: true
 Tag: py3-none-any
 

{onnxtr-0.1.2.dist-info → onnxtr-0.3.0.dist-info}/top_level.txt

@@ -1,2 +1 @@
-docs
 onnxtr