Skip to content

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

feat(rapidocr): support img url as input and optimize vis (#356)

SWHL47418c200549
diff --git a/python/demo.py b/python/demo.py
index c2ee3e9..569bccc 100644
--- a/python/demo.py
+++ b/python/demo.py
@@ -1,29 +1,13 @@
# -*- encoding: utf-8 -*-
# @Author: SWHL
# @Contact: [email protected]
-import cv2
-
from rapidocr import RapidOCR, VisRes
-engine = RapidOCR(params={"Global.with_openvino": True})
+engine = RapidOCR()
vis = VisRes()
-image_path = "tests/test_files/ch_en_num.jpg"
-with open(image_path, "rb") as f:
- img = f.read()
-
-result = engine(img, return_word_box=True)
+img_url = "https://img1.baidu.com/it/u=3619974146,1266987475&fm=253&fmt=auto&app=138&f=JPEG?w=500&h=516"
+result = engine(img_url)
print(result)
-print(result.elapse)
-
-boxes = result.boxes
-txts = result.txts
-scores = result.scores
-
-vis_img = vis(img, result.boxes, result.txts, result.scores)
-cv2.imwrite("vis.png", vis_img)
-words_results = result.word_results
-words, words_scores, words_boxes = list(zip(*words_results))
-vis_img = vis(img, words_boxes, words, words_scores)
-cv2.imwrite("vis_single.png", vis_img)
+result.vis()
diff --git a/python/rapidocr/main.py b/python/rapidocr/main.py
index 83a728c..d863bc5 100644
--- a/python/rapidocr/main.py
+++ b/python/rapidocr/main.py
@@ -99,11 +99,11 @@ class RapidOCR:
self.text_det.postprocess_op.box_thresh = box_thresh
self.text_det.postprocess_op.unclip_ratio = unclip_ratio
- img = self.load_img(img_content)
+ ori_img = self.load_img(img_content)
- raw_h, raw_w = img.shape[:2]
+ raw_h, raw_w = ori_img.shape[:2]
op_record = {}
- img, ratio_h, ratio_w = self.preprocess(img)
+ img, ratio_h, ratio_w = self.preprocess(ori_img)
op_record["preprocess"] = {"ratio_h": ratio_h, "ratio_w": ratio_w}
det_res, cls_res, rec_res = TextDetOutput(), TextClsOutput(), TextRecOutput()
@@ -148,7 +148,7 @@ class RapidOCR:
det_res.boxes, op_record, raw_h, raw_w
)
- ocr_res = self.get_final_res(det_res, cls_res, rec_res)
+ ocr_res = self.get_final_res(ori_img, det_res, cls_res, rec_res)
return ocr_res
def preprocess(self, img: np.ndarray) -> Tuple[np.ndarray, float, float]:
@@ -262,7 +262,11 @@ class RapidOCR:
return dt_boxes_array
def get_final_res(
- self, det_res: TextDetOutput, cls_res: TextClsOutput, rec_res: TextRecOutput
+ self,
+ ori_img: np.ndarray,
+ det_res: TextDetOutput,
+ cls_res: TextClsOutput,
+ rec_res: TextRecOutput,
) -> Union[TextDetOutput, TextClsOutput, TextRecOutput, RapidOCROutput]:
dt_boxes = det_res.boxes
txt_res = rec_res.txts
@@ -280,6 +284,7 @@ class RapidOCR:
return det_res
ocr_res = RapidOCROutput(
+ img=ori_img,
boxes=det_res.boxes,
txts=rec_res.txts,
scores=rec_res.scores,
diff --git a/python/rapidocr/utils/load_image.py b/python/rapidocr/utils/load_image.py
index f34b549..86ab953 100644
--- a/python/rapidocr/utils/load_image.py
+++ b/python/rapidocr/utils/load_image.py
@@ -7,8 +7,11 @@ from typing import Any, Union
import cv2
import numpy as np
+import requests
from PIL import Image, UnidentifiedImageError
+from .utils import is_url
+
root_dir = Path(__file__).resolve().parent
InputType = Union[str, np.ndarray, bytes, Path, Image.Image]
@@ -30,9 +33,14 @@ class LoadImage:
def load_img(self, img: InputType) -> np.ndarray:
if isinstance(img, (str, Path)):
- self.verify_exist(img)
+ if is_url(img):
+ img = Image.open(requests.get(img, stream=True, timeout=60).raw)
+ else:
+ self.verify_exist(img)
+ img = Image.open(img)
+
try:
- img = self.img_to_ndarray(Image.open(img))
+ img = self.img_to_ndarray(img)
except UnidentifiedImageError as e:
raise LoadImageError(f"cannot identify image file {img}") from e
return img
diff --git a/python/rapidocr/utils/typings.py b/python/rapidocr/utils/typings.py
index 2d3f05c..a843ea3 100644
--- a/python/rapidocr/utils/typings.py
+++ b/python/rapidocr/utils/typings.py
@@ -4,11 +4,18 @@
from dataclasses import dataclass, field
from typing import List, Optional, Tuple, Union
+import cv2
import numpy as np
+from .logger import Logger
+from .vis_res import VisRes
+
+logger = Logger(logger_name=__name__).get_log()
+
@dataclass
class RapidOCROutput:
+ img: Optional[np.ndarray] = None
boxes: Optional[np.ndarray] = None
txts: Optional[Tuple[str]] = None
scores: Optional[Tuple[float]] = None
@@ -44,3 +51,23 @@ class RapidOCROutput:
for box, rec in zip(dt_boxes, rec_res):
final_res.append([box, list(rec)])
return final_res
+
+ def vis(self):
+ if self.img is None or self.boxes is None:
+ logger.warning("No image or boxes to visualize.")
+ return
+
+ vis = VisRes()
+ if all(v is None for v in self.word_results):
+ vis_img = vis(self.img, self.boxes, self.txts, self.scores)
+ cv2.imwrite("vis.png", vis_img)
+ logger.info("Visualization saved as vis.png.")
+ return
+
+ # single word vis
+ words_results = self.word_results
+ words, words_scores, words_boxes = list(zip(*words_results))
+ vis_img = vis(self.img, words_boxes, words, words_scores)
+ cv2.imwrite("vis_single.png", vis_img)
+ logger.info("Single word visualization saved as vis_single.png.")
+ return
diff --git a/python/rapidocr/utils/utils.py b/python/rapidocr/utils/utils.py
index 4fa4830..e4e0fee 100644
--- a/python/rapidocr/utils/utils.py
+++ b/python/rapidocr/utils/utils.py
@@ -4,11 +4,20 @@
import importlib
from pathlib import Path
from typing import Union
+from urllib.parse import urlparse
import requests
from tqdm import tqdm
+def is_url(url: str) -> bool:
+ try:
+ result = urlparse(url)
+ return all([result.scheme, result.netloc])
+ except Exception as e:
+ return False
+
+
def import_package(name, package=None):
try:
module = importlib.import_module(name, package=package)
diff --git a/python/tests/test_files/empty_black.jpg b/python/tests/test_files/empty_black.jpg
new file mode 100644
index 0000000..97ed9f8
Binary files /dev/null and b/python/tests/test_files/empty_black.jpg differ
diff --git a/python/tests/test_main.py b/python/tests/test_main.py
index 1d11859..b338d4b 100644
--- a/python/tests/test_main.py
+++ b/python/tests/test_main.py
@@ -38,6 +38,20 @@ def get_engine(params: Optional[Dict[str, Any]] = None):
return engine
+def test_full_black_img(engine):
+ img_path = tests_dir / "empty_black.jpg"
+ result = engine(img_path)
+ assert result.img is None
+ assert result.boxes is None
+
+
+def test_img_url_input(engine):
+ img_url = "https://github.com/RapidAI/RapidOCR/blob/a9bb7c1f44b6e00556ada90ac588f020d7637c4b/python/tests/test_files/ch_en_num.jpg?raw=true"
+ result = engine(img_url)
+ assert result.txts is not None
+ assert result.txts[0] == "正品促销"
+
+
def test_server_rec():
engine = RapidOCR(params={"Global.lang_rec": "ch_server"})
result = engine(img_path)