Why return_word_box didn't works?

[mkrzywda]

问题描述 / Problem Description

I have a Python script with RapidOCR set to return_word_box: True, but it doesn't work as it still generates bounding boxes per line instead of per word. This is critical for me because RapidOCR is a key component of my project, and if it doesn't work correctly with word-level bounding boxes, I'll have to look for another solution.

运行环境 / Runtime Environment

WSL + OpenVINO

复现代码 / Reproduction Code

from rapidocr import RapidOCR
import time

# Initialize RapidOCR with parameters
config_path = "default_rapidocr.yaml"
engine = RapidOCR(config_path=config_path)

print(engine.config)
img_path = "test.png"

# Measure start time
start_time = time.time()

# Run OCR
result = engine(img_path, use_det=True, use_cls=False, use_rec=True)

# Measure end time and calculate duration
end_time = time.time()
duration_ms = (end_time - start_time) * 1000  # Convert to milliseconds

print("OCR Result:", result)
print(f"Processing time: {duration_ms:.2f} ms")

# Visualize result
result.vis('vis_result-openvino-rapid-ocr.jpg')

default_config.yaml

Global:
    lang_det: "en_mobile" # ch_server
    lang_rec: "en_mobile"
    text_score: 0.5

    use_det: true
    use_cls: false
    use_rec: true

    min_height: 30
    width_height_ratio: 8
    max_side_len: 2000
    min_side_len: 30

    return_word_box: true

    with_onnx: false
    with_openvino: true
    with_paddle: false
    with_torch: false

    font_path: null

EngineConfig:
    onnxruntime:
        intra_op_num_threads: -1
        inter_op_num_threads: -1
        use_cuda: true
        use_dml: false

    openvino:
        inference_num_threads: -1

    paddlepaddle:
        cpu_math_library_num_threads: -1
        use_cuda: false
        gpu_id: 0
        gpu_mem: 500

    torch:
        use_cuda: false
        gpu_id: 0

Det:
    model_path: null
    model_dir: null

    limit_side_len: 736
    limit_type: min
    std: [ 0.5, 0.5, 0.5 ]
    mean: [ 0.5, 0.5, 0.5 ]

    thresh: 0.3
    box_thresh: 0.5
    max_candidates: 1000
    unclip_ratio: 1.6
    use_dilation: true
    score_mode: fast

Cls:
    model_path: null
    model_dir: null

    cls_image_shape: [3, 48, 192]
    cls_batch_num: 6
    cls_thresh: 0.9
    label_list: ['0', '180']

Rec:
    model_dir: null
    dict_url: null
    rec_keys_path: null
    rec_img_shape: [3, 48, 320]
    rec_batch_num: 6

可能解决方案 / Possible solutions

• Handling Bouding Boxes per word instead of whole lline.

[wukonggeo]

@mkrzywda
You can manually pass parameters, such as.

from rapidocr_onnxruntime import RapidOCR

img_file = "test.png"
engine = RapidOCR()
result, elapse = engine(img_file, return_word_box=True)

[mkrzywda]

@wukonggeo

I modified the code to:

result = engine(img_path, use_det=True, use_cls=False, use_rec=True, return_word_box=True)

However, even though I set return_word_box=True in the config, the screenshot shows bounding boxes around individual letters instead of words. How can I fix this to get bounding boxes per word?

Also, why should I use rapidocr_onnxruntime when it doesn't have the .vis method?

[SWHL]

The return_word_box parameter means to return the coordinates of a single character. In Chinese, it is a single character, while in English it is a single letter.

If you want to get the word bbox of English, you could use the v1.4.0 of rapidocr_onnxruntime (See link for details)

[mkrzywda]

Okay, I downgraded to rapidocr_onnxruntime to 1.4.0

So: pip show rapidocr_onnxruntime return:

Name: rapidocr-onnxruntime
Version: 1.4.0
Summary: A cross platform OCR Library based on OnnxRuntime.
Home-page: https://github.com/RapidAI/RapidOCR
Author: SWHL
Author-email: liekkaskono@163.com

and use code from your link:

import cv2

from rapidocr_onnxruntime import RapidOCR, VisRes

engine = RapidOCR()
vis = VisRes()

image_path = "test.png"
img = cv2.imread(image_path)
result, elapse_list = engine(img, return_word_box=True)

(boxes, txts, scores, words_boxes, words, words_scores) = list(zip(*result))
font_path = "resources/fonts/FZYTK.TTF"

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)
cv2.imwrite("vis_single.png", vis_img)

and I recived error:

Traceback (most recent call last):
  File "/home/mkrzywda/rapid-ocr-apporach/script-rapidOCRBBword.py", line 15, in <module>
    (boxes, txts, scores, words_boxes, words, words_scores) = list(zip(*result))
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: not enough values to unpack (expected 6, got 5)

in result I have:

result=[
    [
        [[64.0, 13.0], [131.0, 13.0], [131.0, 37.0], [64.0, 37.0]], 'Ubuntu', 0.9998050332069397,[[[72, 13], [124, 13], [124, 37], [72, 37]]],['Ubuntu']],
        [[[424.0, 12.0], [492.0, 15.0], [491.0, 39.0], [422.0, 35.0]], 'Ubuntu', 0.9995335936546326, [[[431, 12], [483, 14], [482, 39], [429, 36]]], ['Ubuntu']],
        [[[785.0, 11.0], [853.0, 14.0], [852.0, 38.0], [783.0, 34.0]], 'Ubuntu', 0.9985363483428955, [[[792, 11], [844, 13], [843, 38], [790, 35]]], ['Ubuntu']],
        [[[11.0, 60.0], [449.0, 60.0], [449.0, 89.0], [11.0, 89.0]], 'from rapidocr import RapidocR', 0.9813520908355713, [[[15, 60], [442, 60], [442, 89], [15, 89]]], ['from rapidocr import RapidocR']],
        [[[10.0, 89.0], [180.0, 91.0], [180.0, 118.0], [10.0, 116.0]], 'import time', 0.9711170196533203, [[[14, 89], [172, 90], [172, 117], [14, 116]]], ['import time']],
        [[[9.0, 147.0], [568.0, 149.0], [568.0, 176.0], [9.0, 174.0]], '# Initialize RapidocR with parameters', 0.9835678339004517, [[[18, 147], [564, 148], [564, 175], [18, 174]]], ['# Initialize RapidocR with parameters']], 
        [[[8.0, 176.0], [563.0, 174.0], [563.0, 206.0], [8.0, 208.0]], 'config_path = "default_rapidocr.yaml"', 0.9967984557151794, [[[18, 175], [559, 174], [559, 206], [18, 207]]], ['config_path = "default_rapidocr.yaml"']], 
        [[[11.0, 206.0], [640.0, 206.0], [640.0, 235.0], [11.0, 235.0]], 'engine = RapidocR(config_path=config_path)', 0.9848083257675171, [[[15, 206], [638, 206], [638, 235], [15, 235]]], ['engine = RapidocR(config_path=config_path)']],
        [[[13.0, 265.0], [310.0, 265.0], [310.0, 292.0], [13.0, 292.0]], 'print(engine.config)', 0.9994365572929382, [[[17, 265], [306, 265], [306, 292], [17, 292]]], ['print(engine.config)']], 
        [[[9.0, 291.0], [315.0, 293.0], [315.0, 322.0], [9.0, 320.0]], 'img_path = "test.png', 0.9972468614578247, [[[13, 291], [304, 292], [304, 321], [13, 320]]], ['img_path = "test.png']],
        [[[10.0, 323.0], [314.0, 323.0], [314.0, 349.0], [10.0, 349.0]], '# Measure start time', 0.9891046285629272, [[[18, 323], [305, 323], [305, 349], [18, 349]]], ['# Measure start time']],
        [[[10.0, 351.0], [371.0, 351.0], [371.0, 380.0], [10.0, 380.0]], 'start_time = time.time()', 0.9989530444145203, [[[14, 351], [369, 351], [369, 380], [14, 380]]], ['start_time = time.time()']],
        [[[10.0, 409.0], [149.0, 409.0], [149.0, 435.0], [10.0, 435.0]], '# Run OCR', 0.9314685463905334, [[[18, 409], [140, 409], [140, 435], [18, 435]]], ['# Run OCR']], 
        [[[10.0, 438.0], [1029.0, 438.0], [1029.0, 467.0], [10.0, 467.0]], 'result = engine(img_path, use_det=True, use_cls=False, use_rec=True)', 0.9926537871360779, [[[14, 438], [1023, 438], [1022, 467], [14, 467]]], ['result = engine(img_path, use_det=True, use_cls=False, use_rec=True)']], 
        [[[9.0, 496.0], [630.0, 496.0], [630.0, 525.0], [9.0, 525.0]], '# Measure end time and calculate duration', 0.9809231758117676, [[[18, 496], [620, 496], [620, 525], [18, 525]]], ['# Measure end time and calculate duration']],
        [[[10.0, 526.0], [343.0, 526.0], [343.0, 554.0], [10.0, 554.0]], 'end_time = time.time()', 0.9839009642601013, [[[14, 526], [337, 526], [337, 554], [14, 554]]], ['end_time = time.time()']],
        [[[10.0, 554.0], [1077.0, 552.0], [1077.0, 581.0], [10.0, 583.0]], 'duration_ms = (end_time - start_time) * 1ooo # Convert to milliseconds', 0.9866061806678772, [[[14, 553], [1067, 552], [1067, 581], [14, 582]]], ['duration_ms = (end_time - start_time) * 1ooo # Convert to milliseconds']], 
        [[[10.0, 612.0], [317.0, 608.0], [317.0, 638.0], [10.0, 642.0]], 'print("OCR Result:",', 0.9628008008003235, [[[15, 611], [310, 608], [310, 638], [15, 641]]], ['print("OCR Result:",']], 
        [[[321.0, 613.0], [431.0, 613.0], [431.0, 640.0], [321.0, 640.0]], 'result)', 0.9958106875419617, [[[330, 613], [424, 613], [424, 640], [330, 640]]], ['result)']], 
        [[[10.0, 641.0], [715.0, 640.0], [715.0, 669.0], [10.0, 670.0]], 'print(f"Processing time: {duration_ms:.2f} ms")', 0.99371337890625, [[[14, 640], [710, 640], [710, 669], [14, 669]]], ['print(f"Processing time: {duration_ms:.2f} ms")']],
        [[[11.0, 698.0], [284.0, 698.0], [284.0, 724.0], [11.0, 724.0]], '# Visualize result', 0.9719812870025635, [[[15, 698], [275, 698], [275, 724], [15, 724]]], ['# Visualize result']], 
        [[[10.0, 726.0], [713.0, 728.0], [713.0, 758.0], [10.0, 756.0]], "result.vis('vis_result-openvino-rapid-ocr.jpg')", 0.9984481334686279, [[[14, 726], [709, 727], [709, 757], [14, 756]]], ["result.vis('vis_result-openvino-rapid-ocr.jpg')"]],
        [[[56.0, 960.0], [155.0, 960.0], [155.0, 987.0], [56.0, 987.0]], 'INSERT', 0.9944579005241394, [[[60, 960], [141, 960], [141, 987], [60, 987]]], ['INSERT']],
        [[[1598.0, 958.0], [1667.0, 958.0], [1667.0, 990.0], [1598.0, 990.0]], '9,22', 0.99979168176651, [[[1603, 958], [1656, 958], [1656, 990], [1603, 990]]], ['9,22']],
        [[[1801.0, 954.0], [1909.0, 956.0], [1909.0, 980.0], [1801.0, 978.0]], 'Pokaz pulpit', 0.9791045188903809, [[[1809, 954], [1905, 955], [1905, 979], [1809, 978]]], ['Pokaz pulpit']],
        [[[52.0, 1021.0], [72.0, 1021.0], [72.0, 1037.0], [52.0, 1037.0]], '18°', 0.9920897483825684, [[[52, 1021], [70, 1021], [70, 1037], [52, 1037]]], ['18°']],
        [[[1845.0, 1020.0], [1895.0, 1020.0], [1895.0, 1042.0], [1845.0, 1042.0]], '11:38', 0.9739528894424438, [[[1848, 1020], [1889, 1020], [1889, 1042], [1848, 1042]]], ['11:38']],
        [[[1801.0, 1042.0], [1895.0, 1044.0], [1895.0, 1068.0], [1801.0, 1066.0]], '22.04.2025', 0.9998246431350708, [[[1805, 1042], [1889, 1043], [1889, 1067], [1805, 1066]]], ['22.04.2025']]    
]

in elapse_list I have:

elapse_list=[0.7135759360025986, 0.02488851547241211, 1.857579231262207]

[mkrzywda]

So it's still results for Line, but you write that in 1.4.0 it will be per Word.

[wukonggeo]

@mkrzywda You can split character into words one by one, separator str/character using “ ” or ""，then merge all bboxs in a word.

[mkrzywda]

Yes, it will be my last approach if in-built method didn't work. Currently I hope that in-built will be work.

[mkrzywda]

Any updates in this topic?

[SWHL]

You can use the latest rapidocr as the following:

from rapidocr import LangRec, RapidOCR

engine = RapidOCR(
    params={
        "Rec.lang_type": LangRec.EN,
    }
)

img_url = "https://github.com/RapidAI/RapidOCR/blob/main/python/tests/test_files/en.jpg?raw=true"
result = engine(img_url, return_word_box=True)
print(result)

result.vis("vis_result.jpg")