Skip to content

📄 Awesome OCR multiple programing languages toolkits based on ONNX Runtime, OpenVINO, MNN, PaddlePaddle, TensorRT and PyTorch.

chore: update files

SWHL12ea82affebd
diff --git a/python/demo.py b/python/demo.py
index e115f06..4d456d6 100644
--- a/python/demo.py
+++ b/python/demo.py
@@ -19,28 +19,19 @@ with open(image_path, "rb") as f:
img = f.read()
# result, elapse_list = engine(img, use_det=True, use_cls=False, use_rec=False)
-result, elapse_list = engine(img, return_word_box=True)
+result = engine(img, return_word_box=True)
print(result)
-print(elapse_list)
+print(result.elapse)
-# result, elapse = engine(image_path, use_det=False, use_cls=False, use_rec=True)
+boxes = result.boxes
+txts = result.txts
+scores = result.scores
-(boxes, txts, scores, words_boxes, words, words_scores) = list(zip(*result))
font_path = "resources/fonts/FZYTK.TTF"
+vis_img = vis(img, result.boxes, result.txts, result.scores, font_path)
+cv2.imwrite("vis.png", vis_img)
-words_boxes = sum(words_boxes, [])
-words_all = sum(words, [])
-words_scores = sum(words_scores, [])
-vis_img = vis(img, words_boxes, words_all, words_scores, font_path)
+words_results = result.word_results
+words, words_scores, words_boxes = list(zip(*words_results))
+vis_img = vis(img, words_boxes, words, words_scores, font_path)
cv2.imwrite("vis_single.png", vis_img)
-
-# (boxes, txts, scores, words_boxes, words) = list(zip(*result))
-
-# vis_img = vis(img, boxes, txts, scores, font_path)
-# cv2.imwrite("vis.png", vis_img)
-
-# words_boxes = sum(words_boxes, [])
-# words_all = sum(words, [])
-# words_scores = [1.0] * len(words_boxes)
-# vis_img = vis(img, words_boxes, words_all, words_scores, font_path)
-# cv2.imwrite("vis_single.png", vis_img)
diff --git a/python/rapidocr/cal_rec_boxes/main.py b/python/rapidocr/cal_rec_boxes/main.py
index 6d86bd9..dc37119 100644
--- a/python/rapidocr/cal_rec_boxes/main.py
+++ b/python/rapidocr/cal_rec_boxes/main.py
@@ -28,7 +28,7 @@ class CalRecBoxes:
for idx, (img, box) in enumerate(zip(imgs, dt_boxes)):
direction = self.get_box_direction(box)
- rec_txt = rec_res.line_txts[idx]
+ rec_txt = rec_res.txts[idx]
rec_word_info = rec_res.word_results[idx]
h, w = img.shape[:2]
diff --git a/python/rapidocr/ch_ppocr_cls/utils.py b/python/rapidocr/ch_ppocr_cls/utils.py
index c4e96b2..fc82ffb 100644
--- a/python/rapidocr/ch_ppocr_cls/utils.py
+++ b/python/rapidocr/ch_ppocr_cls/utils.py
@@ -23,6 +23,11 @@ class TextClsOutput:
cls_res: Optional[List[Tuple[str, float]]] = None
elapse: Optional[float] = None
+ def __len__(self):
+ if self.img_list is None:
+ return 0
+ return len(self.img_list)
+
class ClsPostProcess:
def __init__(self, label_list: List[str]):
diff --git a/python/rapidocr/ch_ppocr_det/utils.py b/python/rapidocr/ch_ppocr_det/utils.py
index 377e9f2..8210428 100644
--- a/python/rapidocr/ch_ppocr_det/utils.py
+++ b/python/rapidocr/ch_ppocr_det/utils.py
@@ -16,6 +16,11 @@ class TextDetOutput:
scores: Optional[Tuple[float]] = None
elapse: float = 0.0
+ def __len__(self):
+ if self.boxes is None:
+ return 0
+ return len(self.boxes)
+
class DetPreProcess:
def __init__(
diff --git a/python/rapidocr/ch_ppocr_rec/main.py b/python/rapidocr/ch_ppocr_rec/main.py
index a67955c..44ea52b 100644
--- a/python/rapidocr/ch_ppocr_rec/main.py
+++ b/python/rapidocr/ch_ppocr_rec/main.py
@@ -90,8 +90,8 @@ class TextRecognizer:
elapse += time.perf_counter() - start_time
all_line_results, all_word_results = list(zip(*rec_res))
- line_txts, line_scores = list(zip(*all_line_results))
- return TextRecOutput(line_txts, line_scores, all_word_results, elapse)
+ txts, scores = list(zip(*all_line_results))
+ return TextRecOutput(txts, scores, all_word_results, elapse)
def resize_norm_img(self, img: np.ndarray, max_wh_ratio: float) -> np.ndarray:
img_channel, img_height, img_width = self.rec_image_shape
diff --git a/python/rapidocr/ch_ppocr_rec/utils.py b/python/rapidocr/ch_ppocr_rec/utils.py
index 4070593..d69410e 100644
--- a/python/rapidocr/ch_ppocr_rec/utils.py
+++ b/python/rapidocr/ch_ppocr_rec/utils.py
@@ -29,13 +29,18 @@ class TextRecInput:
@dataclass
class TextRecOutput:
- line_txts: Optional[Tuple[str]] = None
- line_scores: Tuple[float] = (1.0,)
+ txts: Optional[Tuple[str]] = None
+ scores: Tuple[float] = (1.0,)
word_results: Tuple[Tuple[str, float, Optional[List[List[int]]]]] = (
("", 1.0, None),
)
elapse: Optional[float] = None
+ def __len__(self):
+ if self.txts is None:
+ return 0
+ return len(self.txts)
+
class CTCLabelDecode:
def __init__(
diff --git a/python/rapidocr/main.py b/python/rapidocr/main.py
index bd2451c..3cceb99 100644
--- a/python/rapidocr/main.py
+++ b/python/rapidocr/main.py
@@ -64,7 +64,7 @@ class RapidOCR:
self.cal_rec_boxes = CalRecBoxes()
- self.return_paddleocr_format = True
+ self.return_paddleocr_format = False
def __call__(
self,
@@ -255,7 +255,7 @@ class RapidOCR:
self, det_res: TextDetOutput, cls_res: TextClsOutput, rec_res: TextRecOutput
) -> Union[TextDetOutput, TextClsOutput, TextRecOutput, RapidOCROutput]:
dt_boxes = det_res.boxes
- txt_res = rec_res.line_txts
+ txt_res = rec_res.txts
if dt_boxes is None and txt_res is None and cls_res.cls_res is not None:
return cls_res
@@ -266,13 +266,13 @@ class RapidOCR:
if dt_boxes is None and txt_res is not None:
return rec_res
- if dt_boxes is not None and rec_res is None:
+ if dt_boxes is not None and txt_res is None:
return det_res
ocr_res = RapidOCROutput(
boxes=det_res.boxes,
- txts=rec_res.line_txts,
- scores=rec_res.line_scores,
+ txts=rec_res.txts,
+ scores=rec_res.scores,
word_results=rec_res.word_results,
elapse_list=[det_res.elapse, cls_res.elapse, rec_res.elapse],
)
diff --git a/python/rapidocr/utils/typings.py b/python/rapidocr/utils/typings.py
index 6847986..0ac02cd 100644
--- a/python/rapidocr/utils/typings.py
+++ b/python/rapidocr/utils/typings.py
@@ -17,13 +17,19 @@ class RapidOCROutput:
elapse: float = field(init=False)
def __post_init__(self):
- self.elapse = sum(self.elapse_list)
+ self.elapse = sum(v for v in self.elapse_list if isinstance(v, float))
+
+ def __len__(self):
+ if self.txts is None:
+ return 0
+ return len(self.txts)
def to_json(self):
pass
def to_paddleocr_format(self):
- """Return format like:
+ """Refer: https://pypi.org/project/paddleocr/2.7.3/
+ Return format like:
[
[[[6.0, 2.0], [322.0, 9.0], [320.0, 104.0], [4.0, 97.0]], ['正品促销', 0.99893]],
[[[70.0, 98.0], [252.0, 98.0], [252.0, 125.0], [70.0, 125.0]], ['大桶装更划算', 0.9843]]
diff --git a/python/tests/test_main.py b/python/tests/test_main.py
index b2fc534..b46c578 100644
--- a/python/tests/test_main.py
+++ b/python/tests/test_main.py
@@ -27,10 +27,10 @@ def test_long_img():
img_url = "https://github.com/RapidAI/RapidOCR/releases/download/v1.1.0/long.jpeg"
img_path = tests_dir / "long.jpeg"
download_file(img_url, save_path=img_path)
- result, _ = engine(img_path)
+ result = engine(img_path)
assert result is not None
- assert len(result) >= 53
+ assert len(result.boxes) >= 53
img_path.unlink()
@@ -53,8 +53,8 @@ def test_ort_dml_warning(caplog):
def test_mode_one_img():
img_path = tests_dir / "issue_170.png"
- result, _ = engine(img_path)
- assert result[0][1] == "TEST"
+ result = engine(img_path)
+ assert result.txts[0] == "TEST"
@pytest.mark.parametrize(
@@ -72,8 +72,8 @@ def test_mode_one_img():
)
def test_transparent_img(img_name: str, gt: str):
img_path = tests_dir / img_name
- result, _ = engine(img_path)
- assert result[0][1] == gt
+ result = engine(img_path)
+ assert result.txts[0] == gt
@pytest.mark.parametrize(
@@ -89,54 +89,53 @@ def test_transparent_img(img_name: str, gt: str):
)
def test_letterbox_like(img_name, gt_len, gt_first_len):
img_path = tests_dir / img_name
- result, _ = engine(img_path)
+ result = engine(img_path)
assert len(result) == gt_len
- assert result[0][1].lower() == gt_first_len.lower()
+ assert result.txts[0].lower() == gt_first_len.lower()
def test_only_det():
- result, _ = engine(img_path, use_det=True, use_cls=False, use_rec=False)
-
+ result = engine(img_path, use_det=True, use_cls=False, use_rec=False)
assert len(result) == 18
def test_only_cls():
img_path = tests_dir / "text_cls.jpg"
- result, _ = engine(img_path, use_det=False, use_cls=True, use_rec=False)
+ result = engine(img_path, use_det=False, use_cls=True, use_rec=False)
assert len(result) == 1
- assert result[0][0] == "0"
+ assert result.cls_res[0][0] == "0"
def test_only_rec():
img_path = tests_dir / "text_rec.jpg"
- result, _ = engine(img_path, use_det=False, use_cls=False, use_rec=True)
+ result = engine(img_path, use_det=False, use_cls=False, use_rec=True)
assert len(result) == 1
- assert result[0][0] == "韩国小馆"
+ assert result.txts[0] == "韩国小馆"
def test_det_rec():
- result, _ = engine(img_path, use_det=True, use_cls=False, use_rec=True)
- assert result[0][1] == "正品促销"
+ result = engine(img_path, use_det=True, use_cls=False, use_rec=True)
assert len(result) == 18
+ assert result.txts[0] == "正品促销"
def test_cls_rec():
img_path = tests_dir / "text_cls.jpg"
- result, _ = engine(img_path, use_det=False, use_cls=True, use_rec=True)
+ result = engine(img_path, use_det=False, use_cls=True, use_rec=True)
assert result is not None
assert len(result) == 1
- assert result[0][0] == "韩国小馆"
+ assert result.txts[0] == "韩国小馆"
def test_det_cls_rec():
img = cv2.imread(str(img_path))
- result, _ = engine(img)
+ result = engine(img)
assert result is not None
- assert result[0][1] == "正品促销"
assert len(result) == 18
+ assert result.txts[0] == "正品促销"
def test_empty():
@@ -149,41 +148,40 @@ def test_empty():
def test_zeros():
img = np.zeros([640, 640, 3], np.uint8)
- result, _ = engine(img)
- assert result is None
+ result = engine(img)
+ assert result.boxes is None
def test_input_str():
- result, _ = engine(str(img_path))
- assert result[0][1] == "正品促销"
+ result = engine(str(img_path))
assert len(result) == 18
+ assert result.txts[0] == "正品促销"
def test_input_bytes():
with open(img_path, "rb") as f:
- result, _ = engine(f.read())
- assert result[0][1] == "正品促销"
+ result = engine(f.read())
assert len(result) == 18
+ assert result.txts[0] == "正品促销"
def test_input_path():
- result, _ = engine(img_path)
- assert result[0][1] == "正品促销"
+ result = engine(img_path)
assert len(result) == 18
+ assert result.txts[0] == "正品促销"
def test_input_parameters():
img_path = tests_dir / "ch_en_num.jpg"
engine = RapidOCR(text_score=1)
- result, _ = engine(img_path)
-
- assert result is None
+ result = engine(img_path)
+ assert result.boxes is None
def test_input_det_parameters():
with pytest.raises(FileNotFoundError) as exc_info:
engine = RapidOCR(det_model_path="1.onnx")
- result, _ = engine(img_path)
+ result = engine(img_path)
raise FileNotFoundError()
assert exc_info.type is FileNotFoundError
@@ -191,7 +189,7 @@ def test_input_det_parameters():
def test_input_cls_parameters():
with pytest.raises(FileNotFoundError) as exc_info:
engine = RapidOCR(cls_model_path="1.onnx")
- result, _ = engine(img_path)
+ result = engine(img_path)
raise FileNotFoundError()
assert exc_info.type is FileNotFoundError
@@ -199,7 +197,7 @@ def test_input_cls_parameters():
def test_input_rec_parameters():
with pytest.raises(FileNotFoundError) as exc_info:
engine = RapidOCR(rec_model_path="1.onnx")
- result, _ = engine(img_path)
+ result = engine(img_path)
raise FileNotFoundError()
assert exc_info.type is FileNotFoundError
@@ -207,11 +205,11 @@ def test_input_rec_parameters():
def test_input_three_ndim_two_channel():
img_npy = tests_dir / "two_dim_image.npy"
image_array = np.load(str(img_npy))
- result, _ = engine(image_array)
+ result = engine(image_array)
assert result is not None
assert len(result) == 1
- assert result[0][1] == "TREND PLOT REPORT"
+ assert result.txts[0] == "TREND PLOT REPORT"
def test_input_three_ndim_one_channel():
@@ -220,30 +218,59 @@ def test_input_three_ndim_one_channel():
img = img[:, :, 0]
img = img[..., None] # (H, W, 1)
- result, _ = engine(img)
-
+ result = engine(img)
assert len(result) >= 17
[email protected](reason="no way of currently testing this")
@pytest.mark.parametrize(
"img_name,words",
[
(
"black_font_color_transparent.png",
- ["我", "是", "中", "国", "人"],
+ ("我", "是", "中", "国", "人"),
),
(
"text_vertical_words.png",
- ["已", "取", "之", "時", "不", "參", "一", "人", "見", "而"],
+ (
+ "已",
+ "取",
+ "之",
+ "時",
+ "不",
+ "參",
+ "一",
+ "人",
+ "見",
+ "而",
+ "是",
+ "非",
+ "不",
+ "得",
+ "問",
+ "之",
+ "人",
+ "要",
+ "取",
+ "之",
+ "有",
+ "是",
+ "是",
+ "非",
+ "非",
+ "之",
+ "士",
+ "師",
+ "也",
+ ),
),
(
"issue_170.png",
- ["T", "E", "S", "T"],
+ ("T", "E", "S", "T"),
),
],
)
def test_word_ocr(img_name: str, words: List[str]):
img_path = tests_dir / img_name
- result, _ = engine(img_path, return_word_box=True)
- assert result[0][4] == words
+ result = engine(img_path, return_word_box=True)
+ txts, _, _ = list(zip(*result.word_results))
+ assert txts == words