paddle-ocr-rs 0.6.1

Use Rust to call Paddle OCR models via ONNX Runtime for image text recognition.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159

[简体中文](./docs/README_zh-Hans.md)

## paddle-ocr-rs


Use Rust to call Paddle OCR models via ONNX Runtime for image text recognition.

### Example


```rust
use crate::{ocr_error::OcrError, ocr_lite::OcrLite};

fn run_test() -> Result<(), OcrError> {
    let mut ocr = OcrLite::new();
    ocr.init_models(
        "./models/ch_PP-OCRv5_mobile_det.onnx",
        "./models/ch_ppocr_mobile_v2.0_cls_infer.onnx",
        "./models/ch_PP-OCRv5_rec_mobile_infer.onnx",
        2,
    )?;

    println!("===test_1===");
    let res = ocr.detect_from_path(
        "./docs/test_images/test_1.png",
        50,
        1024,
        0.5,
        0.3,
        1.6,
        false,
        false,
    )?;
    res.text_blocks.iter().for_each(|item| {
        println!("text: {} score: {}", item.text, item.text_score);
    });

    println!("===test_2===");
    let res = ocr.detect_from_path(
        "./docs/test_images/test_2.png",
        50,
        1024,
        0.5,
        0.3,
        1.6,
        false,
        false,
    )?;
    res.text_blocks.iter().for_each(|item| {
        println!("text: {} score: {}", item.text, item.text_score);
    });

    // 通过 image 读取图片
    println!("===test_3===");
    let test_three_img = image::open("./docs/test_images/test_3.png")
        .unwrap()
        .to_rgb8();
    let res = ocr.detect(&test_three_img, 50, 1024, 0.5, 0.3, 1.6, true, false)?;
    res.text_blocks.iter().for_each(|item| {
        println!("text: {} score: {}", item.text, item.text_score);
    });

    Ok(())
}

// 某些情况下角度纠正会得出错误结果，支持角度纠正回退，当角度纠正后的文本识别得分低于指定值（或为 NaN）时，将使用进行角度纠正前的图片进行识别
fn run_test_angle_rollback() -> Result<(), OcrError> {
    let mut ocr = OcrLite::new();
    ocr.init_models(
        "./models/ch_PP-OCRv4_det_infer.onnx",
        "./models/ch_ppocr_mobile_v2.0_cls_infer.onnx",
        "./models/ch_PP-OCRv4_rec_infer.onnx",
        2,
    )?;


    println!("===test_angle_ori===");
    let test_img = image::open("./docs/test_images/test_4.png")
        .unwrap()
        .to_rgb8();
    let res = ocr.detect(&test_img, 50, 1024, 0.5, 0.3, 1.6, true, false)?;
    res.text_blocks.iter().for_each(|item| {
        println!("text: {} score: {}", item.text, item.text_score);
    });


    println!("===test_angle_rollback===");
    let test_img = image::open("./docs/test_images/test_4.png")
        .unwrap()
        .to_rgb8();
    let res =
        ocr.detect_angle_rollback(&test_img, 50, 1024, 0.5, 0.3, 1.6, true, false, 0.8)?;
    res.text_blocks.iter().for_each(|item| {
        println!("text: {} score: {}", item.text, item.text_score);
    });
    Ok(())
}
```

### Reference Development Environment


| Dependency | Version                       |
| ---------- | ----------------------------- |
| rustc      | 1.84.1 (e71f9a9a9 2025-01-27) |
| cargo      | 1.84.1 (66221abde 2024-11-19) |
| OS         | Windows 11 24H2               |
| Paddle OCR | 4                             |

### Documentation


[Error Handling](/docs/error/index.md)

[Static Linking Reference](/docs/staticLinking/index.md)

### Model Source


[RapidOCR Docs](https://rapidai.github.io/RapidOCRDocs/main/model_list/)

### Related Notes


The code is referenced from [RapidOcrOnnx](https://github.com/RapidAI/RapidOcrOnnx), and has replaced OpenCV with image and imageproc libraries for image-related implementations.

### Results Demonstration


#### test_1.png


![test_1](/docs/test_images/test_1.png)

```bash
text: 使用Rust 通过ONNX Runtime 调用 Paddle OCR 模型进行图片文字识别。 score: 0.95269924
text: paddle-ocr-rs score: 0.9979071
```

#### test_2.png


![test_2](/docs/test_images/test_2.png)

```bash
text: 母婴用品连锁 score: 0.99713486
```

#### test_3.png


![test_3](/docs/test_images/test_3.png)

#### Output Preview


```bash
text: salta sobre o cao preguicoso. score: 0.9794339
text: perezoso. A raposa marrom rapida score: 0.9970329
text: marron rapido salta sobre el perro score: 0.9995695
text: salta sopra il cane pigro. El zorro score: 0.99923337
text: paresseux. La volpe marrone rapida score: 0.9991456
text: 《rapide> saute par-dessus le chien score: 0.9685502
text: uber den faulen Hund. Le renard brun score: 0.988613
text: Der ,schnelle" braune Fuchs springt score: 0.97560924
text: from aspammer@website.com is spam. score: 0.98167914
text: & duck/goose, as 12.5% of E-mail score: 0.98472834
text: Over the $43,456.78 <lazy> #90 dog score: 0.9847551
text: The (quick) [brown] {fox} jumps! score: 0.98300403
```