Skip to content

DevGuide.md

Latest commit

 

History

History
958 lines (811 loc) · 33.8 KB

DevGuide.md

File metadata and controls

958 lines (811 loc) · 33.8 KB

English | 한국어 | 日本語 | Tiếng Việt

Application Development Guide

Table of Contents

1. Entry Points

The prototype for all functions is as follows:

#ifdef WIN32
#define TS_ANPR_ENTRY extern "C" __declspec(dllexport) const char* WINAPI
#else
#define TS_ANPR_ENTRY extern "C" const char*
#endif

To avoid being verbose, the following will be denoted as TS_ANPR_ENTRY.

1.1. anpr_initialize

Initializes the library. This function must be called once before using any other functions in the library.

TS_ANPR_ENTRY anpr_initialize(const char* mode); // [IN] Sets the library operation mode

Parameters:

  • mode
    • Used to specify the library's operation mode (the purpose has been expanded from the previous outputFormat)
    • Multiple settings can be specified, separated by a semicolon (;) (e.g., json;sync) 
        # Python Example
  # Multiple settings
  err = anpr_initialize(b'json;sync')
  • Available options:
    • outputFormat:
      • Specifies the output data format
      • Supported formats: text, json, yaml, xml, csv (default: text)
      • You can simply use text or json without specifying outputFormat
    • sync:
      • Runs in synchronous mode (processes requests in the order they are called with thread locking)
      • If your application creates threads dynamically (not using a fixed-size thread pool) and you encounter the following error codes, consider using this option:
        • 103: Too many workers – The number of library call threads exceeds the limit (maximum 256)
        • 104: Resource exhausted – No more resources can be allocated
      • This mode eliminates the need for complex asynchronous thread management, but since it uses thread locks, parallel processing performance may be reduced
      • Can be specified as sync=true or sync=false, or simply as sync (if not specified, the default is sync=false) 
          # Python Example
  # Runs in synchronous mode
  err = anpr_initialize(b'sync')
    • minChar:
      • Specifies the minimum number of characters required for successful license plate recognition
      • If minChar is not specified, is 0, negative, or not a number, the default value 4 is applied
        • Correct example: minChar=5 (recognition succeeds only if at least 5 characters are detected) 
            # Python Example
  # Set number recognition to succeed only if there are 5 or more characters
  err = anpr_initialize(b'minChar=5')
        • Incorrect examples: minChar=0, minChar=-10, minChar=Two (the specified value is ignored and the default 4 is used)
    • country:
      • Specifies the country for the license plate recognition model.
        • Example: country=JP (set for Japanese license plates)
      • Available country codes:
        • KR: Korean license plates
        • JP: Japanese license plates
        • VN: Vietnamese license plates 
          # Python example
# How to specify country
err = anpr_initialize(b'country=KR')  # Korean license plates (default)
err = anpr_initialize(b'country=JP')  # Japanese license plates
err = anpr_initialize(b'country=VN')  # Vietnamese license plates
      • For the free trial, you can specify the country code in the application, but for the commercial license, the supported country is determined by the purchased license.
    • symbol: (For Japanese license plates only)
      • Specifies how to output the · and - characters included in Japanese license plates
        • none: Do not output either · or - (default value)
        • zero: Output · as 0, do not output -
        • dot: Output · as is, do not output -
        • full: Output both · and - as is 
            # Python Example
  # Output both · and - as is
  err = anpr_initialize(b'symbol=full')

Return value:

  • If successful, returns an empty NULL-terminated string (0x00).
  • If an error occurs, returns an error message as a string in the data format specified by the outputFormat in mode (UTF-8 encoded).

1.2. anpr_read_file

Recognizes vehicle license plates from an image file.

TS_ANPR_ENTRY anpr_read_file(
  const char* imgFileName,  // [IN] Input image file name
  const char* outputFormat, // [IN] Output data format
  const char* options);     // [IN] Feature options

Parameters:

  • imgFileName:
    • Input image file name (UTF-8 encoded)
    • Supported image file formats: bmp, jpg, png, pnm, pbm, pgm, ppm, jfif, webp
  • outputFormat:
    • Output data format
    • Supported formats: text, json, yaml, xml, csv (default: text)
    • When using object detection options, text output will be provided in csv format.
  • options:
    • Specifies options for the license plate recognition algorithm. (See: 2. Options)

Return value:

  • Returns the recognition result as a string in the data format specified by outputFormat (UTF-8 encoded).

  • Depending on the options used (license plate recognition or object detection), the outputFormat may differ. (See: 3. Output Format)

      # Python Example
  result = anpr_read_file(b'input-image.jpg', b'json', b'vms')
  if len(result) > 0:
      print(result.decode('utf8'))

Remarks:

1.3. anpr_read_pixels

Recognizes vehicle license plates from the memory buffer of a loaded image. Since TS-ANPR v2.3.0, encoded image buffers are supported.

TS_ANPR_ENTRY anpr_read_pixels(
  const unsigned char* pixels,  // [IN] Starting address of the image pixels
  const unsigned long width,    // [IN] Number of pixels in image width
  const unsigned long height,   // [IN] Number of pixels in image height
  const long stride,            // [IN] Number of bytes per image line
  const char* pixelFormat,      // [IN] Image pixel format
  const char* outputFormat,     // [IN] Output data format
  const char* options);         // [IN] Feature options

Parameters:

  • pixels:
    • Starting address of the image pixels
  • width:
    • Number of pixels in the image width
    • For encoded images, this is the total number of bytes
  • height:
    • Number of pixels in the image height
    • Not used for encoded images (set to default value 0)
  • stride:
    • Number of bytes per image line (if 0, it is assumed there is no padding and calculated automatically)
    • Not used for encoded images (set to default value 0)
  • pixelFormat:
    • Image pixel format
    • Supported pixel formats:
      • GRAY: Grayscale image (8bpp)
      • BGRA: BGRA (32bpp)
      • RGBA: RGBA (32bpp)
      • RGB: RGB (24bpp)
      • BGR: BGR (24bpp)
      • BGR555: BGR (16bpp)
      • BGR565: BGR (16bpp)
      • HSV: HSV (32bpp)
      • YCrCb: YUV444 (32bpp)
      • I420: YUV420 (12bpp)
      • YV12: YUV420 (12bpp)
      • IYUV: YUV420 (12bpp)
      • NV12: YUV420 (12bpp)
      • NV21: YUV420 (12bpp)
    • Supported image encoding formats:
      • bmp, jpg, jpeg, png, pnm, pbm, pgm, ppm, jfif, webp
      • If set to encoded, the image format is automatically detected
  • outputFormat: (Same as anpr_read_file)
  • options: (Same as anpr_read_file)

Return value: (Same as anpr_read_file) Remarks: (Same as anpr_read_file)

  # Python Examples

  # Input a video frame
  ret, frame = capture.read()
  height = frame.shape[0]
  width = frame.shape[1]
  result = anpr_read_pixels(bytes(frame), width, height, 0, b'BGR', b'json', b'vms')
  if len(result) > 0:
      print(result.decode('utf8'))

  # Input an image from the website
  response = requests.get('https://example.com/image.jpg')
  if response.status_code == 200:
      result = anpr_read_pixels(response.content, 0, 0, 0, b'encoded', b'json', b'vms')
      if len(result) > 0:
          print(result.decode('utf8'))

2. Options

  • Various options can be specified to customize the license plate recognition algorithm for your environment.

2.1. Supported Options by License Type

Character Description Applicable License
v Determines whether a license plate is attached to a car All
b Determines whether a license plate is attached to a motorcycle (Vietnamese license plates only) All
m Recognizes all license plates in multiple vehicles Object Detection, Pro, Server
s Recognizes license plates from vehicles at any 360° angle (surround recognition) Object Detection, Pro, Server
d Object detection (for load/unload status) All
r Recognizes license plates of detected vehicles All
i Sets Region of Interest (RoI) All
x Sets Region of Uninterest (RoU) All
a Sets minimum license plate size All
  • Depending on the license type, features and the number of vehicles that can be recognized simultaneously may vary. (See: TS-ANPR Engine)

2.2. Option Selection Guide

In the flowchart below, follow the conditions from Start to Done and collect the characters in the green circles you pass through. Use these collected characters as the option value (the order of the characters does not matter). For example, if your path is Start → v → s → m → Done, the option value would be "vsm".

flowchart TD

  start[/Start/]-->attach{{License plate<br/>attachment status?}}
  attach-->|Don't care|multi
  attach-->|Only license plates<br/>attached to vehicles|v((v)):::opt

  v-->angle{{vehicle angle?}}
  angle-->|Only upright vehicles|multi{{Multi-recognition?}}
  angle-->|All 360° angles|s((s)):::opt
  s-->multi
  multi-->|Yes|m((m)):::opt
  multi-->|No|done[/Done/]
  m-->done

  start-->d((d)):::opt
  d-->angle2{{vehicle angle?}}
  angle2-->|Only upright vehicles|multi2{{Multi-recognition?}}
  angle2-->|All 360° angles|s2((s)):::opt
  s2-->multi2
  multi2-->|Yes|m2((m)):::opt
  multi2-->|No|read
  m2-->read{{License Plate Recognition?}}

  read-->|Yes|r((r)):::opt
  read-->|No|done
  r-->done


  subgraph "License Plate Recognition"
    attach
    v
    angle
    multi
    m
    s
  end

  subgraph "Object detection"
    read
    angle2
    multi2
    s2
    m2
    d
    r
  end

  classDef opt fill:#6F6,stroke:#CCC,stroke-width:4px,padding:10px,font-weight:bold
Loading

  • Examples of License Plate Recognition Options

    Options Description
    Recognizes one license plate among all plates
    v Recognizes one license plate attached to a vehicle
    b Recognizes one license plate attached to a motorcycle (Vietnamese license plates only)
    m Recognizes all license plates (multi-recognition)
    vm Recognizes all license plates attached to vehicles (multi-recognition)
    vs Recognizes one license plate attached to a vehicle, 360° surround recognition
    vsm Recognizes all license plates attached to vehicles, 360° surround recognition (multi-recognition)

  • Examples of Object Detection Options

    Options Description
    d Single object detection (does not recognize vehicle license plates)
    dr Single object detection with vehicle license plate recognition
    ds Single object detection, 360° surround detection (does not recognize vehicle license plates)
    dsr Single object detection, 360° surround detection with vehicle license plate recognition
    dm Multi-object detection (does not recognize vehicle license plates)
    dmr Multi-object detection with vehicle license plate recognition (does not recognize other objects)
    dms Multi-object detection, 360° surround detection (does not recognize vehicle license plates)
    dmsr Multi-object detection, 360° surround detection with vehicle license plate recognition

2.3. Setting Region of Interest (RoI) / Region of Uninterest (RoU)

Region of Interest (RoI)

  • The region of interest is defined as a polygon by connecting the pixel coordinates of each vertex.
    • After the character i, list the pixel coordinates of each vertex in the order of x, y, separated by commas.
  • Multiple regions of interest can be set, and they can be combined with other license plate recognition options.

Region of Uninterest (RoU)

  • The region of uninterest is defined as a polygon by connecting the pixel coordinates of each vertex.
    • After the character x, list the pixel coordinates of each vertex in the order of x, y, separated by commas.
  • Multiple regions of uninterest can be set, and they can be combined with other license plate recognition options.

Note: If neither a region of interest nor a region of uninterest is specified, license plate recognition will be performed on the entire image. If both a region of interest and a region of uninterest are specified and there are overlapping areas, the region of uninterest takes precedence. Therefore, any license plates in the overlapping area will be ignored.

# Python Example
# Given the following vertex pixel coordinates for each polygonal region in the image above:
# RoI = [(810,64), (939,1182), (1486,1182), (1149,571), (879,124), (839,64)]
# RoI2 = [(771,67), (479,1182), (1793,1182), (801,67)]
# RoU = [(851,70), (923,134), (1753,1182), (1789,1182), (1789,250), (1176,87), (946,68)]

# Setting a single region of interest (RoI)
result = anpr_read_file(
  b'roi.jpg',    # Input file name
  b'text',       # Output format
  b'vi810,64,939,1182,1486,1182,1149,571,879,124,839,64') # Use the 'v' option to recognize only license plates attached to vehicles

# Setting two regions of interest (RoI)
result = anpr_read_file(
  b'roi.jpg',    # Input file name
  b'text',       # Output format
  b'vi810,64,939,1182,1486,1182,1149,571,879,124,839,64i771,67,479,1182,1793,1182,801,67')

# Setting one region of interest and one region of uninterest (RoU)
result = anpr_read_file(
  b'roi.jpg',    # Input file name
  b'text',       # Output format
  b'vi810,64,939,1182,1486,1182,1149,571,879,124,839,64x851,70,923,134,1753,1182,1789,1182,1789,250,1176,87,946,68')

# Setting a single region of uninterest (RoU)
result = anpr_read_file(
  b'roi.jpg',    # Input file name
  b'text',       # Output format
  b'vx851,70,923,134,1753,1182,1789,1182,1789,250,1176,87,946,68')

2.4. Setting Minimum License Plate Size

  • The minimum license plate size is specified after the character a as the area of the license plate bounding box (width x height in pixels). For example, if the width is 100 pixels and the height is 40 pixels, enter a4000.

  • If a minimum license plate size is set, any license plate smaller than this will be ignored.

    # Python Example
# In the image, the license plate size is 156 x 154 = 24,024
result = anpr_read_file(
  b'test.jpg',   # Input file name
  b'text',       # Output format
  b'vi7,580, 829,293, 1910,325, 1798,1077, 0,1077a24024'  # Specify RoI and minimum license plate size
)

3. Output Data Formats

The output data is divided into two formats: object detection (when d is included in the options) and vehicle license plate recognition (when d is not included).

3.1. Vehicle License Plate Recognition Results

3.1.1. text

Only the vehicle license plate text is output. If there are multiple license plates, they are separated by the line break character CR (0x0d).

01가2345
67나8901

If no license plate is recognized, an empty text (NULL terminated string (0x00)) is output.

If an error is returned, it is output in the following text format:

error: (1) Invalid parameters
3.1.2. csv

Outputs the vehicle license plate and its attributes in csv format. Each recognized license plate is represented by one line, and each column is separated by a comma (,).

01가2345,1217,2083,92,175,12.45,0.75,0.83,0.20,ev
67나8901,1108,1317,67,217,12.45,0.76,0.89,0.10,

The meaning of each column is as follows:

Column Description Notes
1 Vehicle license plate number text
2 Top-left x coordinate of license plate area.x
3 Top-left y coordinate of license plate area.y
4 License plate width area.width
5 License plate height area.height
6 License plate angle area.angle
7 OCR confidence conf.ocr
8 Plate recognition confidence conf.plate
9 OCR elapsed time (seconds) elapsed
10(1) Eco-friendly electric vehicle status attr.ev
  • (1) (For Korean license plates only) If the vehicle is an eco-friendly electric vehicle, ev is output; otherwise, the field is left blank.

If no license plate is recognized, an empty text (NULL terminated string (0x00)) is output.

If an error is returned, it is output in the following text format:

error,1,Invalid parameters

3.1.3. json

Outputs the vehicle license plate and its attributes in json format.

[
  {
    // First license plate
    text: "01가2345", // Vehicle license plate number
    area: {
      // License plate area (in pixels)
      x: 1217, // Top-left x coordinate
      y: 2083, // Top-left y coordinate
      width: 92, // Width
      height: 175, // Height
      angle: 12.45, // Tilt angle (degrees)
    },
    attrs: {
      // License plate attributes
      ev: true, // Eco-friendly electric vehicle status
    },
    ev: true, // deprecated (moved to attrs.ev, will be removed in future versions)
    conf: {
      // Confidence (range: 0 ~ 1)
      ocr: 0.75, // OCR confidence
      plate: 0.84, // Plate recognition confidence
    },
    elapsed: 0.27, // Elapsed time (in seconds)
  },
  {
    // Second license plate
    text: "67나8901",
    area: {
      x: 1108,
      y: 1317,
      width: 67,
      height: 217,
      angle: 12.45,
    },
    attrs: {
      ev: false,
    },
    ev: false,
    conf: {
      ocr: 0.76,
      plate: 0.89,
    },
    elapsed: 0.14,
  },
];

If no license plate is recognized, empty data is output as follows:

[];

If an error is returned, it is output in the following json format:

{
  "error": {
    "code": 1,
    "message": "Invalid parameters"
  }
}

3.1.4. yaml

Outputs the vehicle license plate and its attributes in yaml format.

- text: 01가2345 # First license plate, vehicle number
  area: # License plate area (in pixels)
    x: 1217 # Top-left x coordinate
    y: 2083 # Top-left y coordinate
    width: 92 # Width
    height: 175 # Height
    angle: 12.45 # Tilt angle (degrees)
  conf: # Confidence (range: 0 ~ 1)
    ocr: 0.75 # OCR confidence
    plate: 0.83 # Plate recognition confidence
  attrs: # License plate attributes
    ev: true # Eco-friendly electric vehicle status
  ev: true # deprecated (moved to attrs.ev, will be removed in future versions)
  elapsed: 0.20 # Elapsed time (seconds)
- text: 67나8901 # Second license plate
  area:
    x: 1108
    y: 1317
    width: 67
    height: 217
    angle: 12.45
  conf:
    ocr: 0.76
    plate: 0.89
  ev: false
  elapsed: 0.10

If no license plate is recognized, empty data is output as follows:



If an error is returned, it is output in the following yaml format:


error
  code: 1
  message: Invalid parameters


3.1.5. xml


Outputs the vehicle license plate and its attributes in xml format.


<?xml version="1.0" encoding="utf-8"?>
<data>
  <!-- First license plate
    text: vehicle license plate number
    ev: deprecated (moved to attrs.ev, will be removed in future versions)
    elapsed: elapsed time (seconds)
  -->
  <license-plate text="01가2345" ev="true" elapsed="0.20">
    <!-- License plate attributes
      ev: eco-friendly electric vehicle status
    -->
    <attrs ev="true"/>
    <!-- License plate area (in pixels)
      x: top-left x coordinate
      y: top-left y coordinate
      width: width
      height: height
      angle: tilt angle (degrees)
    -->
    <area x="1217" y="2083" width="92" height="175" angle="12.45"/>
    <!-- Confidence (range: 0 ~ 1)
      ocr: OCR confidence
      plate: plate recognition confidence
    -->
    <conf ocr="0.75" plate="0.83"/>
  </license-plate>
  <!-- Second license plate -->
  <license-plate text="67나8901" ev="false" elapsed="0.11">
    <attrs ev="false"/>
    <area x="1108" y="1317" width="67" height="217"/>
    <conf ocr="0.76" plate="0.89"/>
  </license-plate>
</data>


If no license plate is recognized, empty data is output as follows:


<?xml version="1.0" encoding="utf-8"?>
<data />


If an error is returned, it is output in the following xml format:


<?xml version="1.0" encoding="utf-8"?>
<error code="1" message="Invalid parameters" />


3.2. Object Detection Results


Detects pre-trained objects in the image.
A distinguishing feature is that it can recognize distorted objects in original images taken with a 360° fisheye lens camera.
Currently supported object classes are as follows:






class
Name






car
Car




motorcycle
Motorcycle






3.2.1. csv, text


Outputs object detection results in csv format text.
If the output format is specified as text, it is also output in csv format.


Each detected object is represented by one line, and each column is separated by a comma (,).


car,2171,2281,396,521,0.9679,0.2886,51조8969,2420,2295,110,81,147.5933,0.9005,0.7864,0.3913,ev
car,264,2266,433,543,0.9706,0.2886,41노7656,315,2281,103,81,211.3135,0.9160,0.8299,0.4189,
car,777,0,579,403,0.9716,0.2886 // When there is no license plate information


The meaning of each column is as follows:






Column
Description
Notes






1
Class
class




2
x pixel coordinate
area.x




3
y pixel coordinate
area.y




4
Width
area.width




5
Height
area.height




6
Confidence
conf




7
Elapsed time (seconds)
elapsed




8(1)
License plate number
licensePlate.text




9
Top-left x coordinate of license plate
licensePlate.area.x




10
Top-left y coordinate of license plate
licensePlate.area.y




11
License plate width
licensePlate.area.width




12
License plate height
licensePlate.area.height




13
License plate angle
licensePlate.area.angle




14
OCR confidence
licensePlate.conf.ocr




15
Plate recognition confidence
licensePlate.conf.plate




16
OCR elapsed time (seconds)
licensePlate.elapsed




17(2)
Eco-friendly electric vehicle status
licensePlate.attrs.ev






    

  • (1) Columns from 8 onward are output only if the r option is used and a license plate is recognized; otherwise, all subsequent columns are omitted.
    • 

  • (2) (For Korean license plates only) Column 17 is output as ev if the vehicle is recognized as an eco-friendly electric vehicle; otherwise, it is left blank.
    • 



If no object is detected, an empty text (NULL terminated string (0x00)) is output.


If an error is returned, it is output in the following text format:


error,1,Invalid parameters


3.2.2. json


Outputs object detection results in json format.


 [
  {                           // First object
    "class": "car",           // Object class
    "area": {                 // Object area (in pixels)
      "x": 2171,              // Top-left x coordinate
      "y": 2281,              // Top-left y coordinate
      "width": 396,           // Width
      "height": 521           // Height
    },
    "conf": 0.9679,           // Object detection confidence (range: 0 ~ 1)
    "elapsed": 0.2513,        // Elapsed time (seconds)
    "licensePlate": [         // License plate(s)
      {
        "text": "51조8969"    // License plate number
        "area": {             // License plate area (in pixels)
          "x": 2420,          // Top-left x coordinate
          "y": 2295           // Top-left y coordinate
          "width": 110,       // Width
          "height": 81,       // Height
          "angle": 147.5933   // Tilt angle (degrees)
        },
        "attrs": {            // License plate attributes
          "ev": true          // Eco-friendly electric vehicle status
        },
        "conf": {             // Confidence (range: 0 ~ 1)
          "ocr": 0.9005,      // OCR confidence
          "plate": 0.7864     // Plate recognition confidence
        },
        "elapsed": 0.3525,    // Elapsed time (seconds)
      }
    ]
  },
  {                           // Second object
    "class": "car",
    "area": {
      "x": 264,
      "y": 2266,
      "width": 433,
      "height": 543
    },
    "conf": 0.9706,
    "elapsed": 0.2513,
    "licensePlate": [
      {
        "text": "41노7656"
        "area": {
          "x": 315,
          "y": 2281,
          "width": 103,
          "height": 81,
          "angle": 211.3135
        },
        "attrs": {
          "ev": false
        },
        "conf": {
          "ocr": 0.916,
          "plate": 0.8299
        },
        "elapsed": 0.4402
      }
    ]
  },
  {                           // Third object (when there is no license plate information)
    "class": "car",
    "area": {
      "x": 777,
      "y": 0
      "height": 403,
      "width": 579,
    },
    "conf": 0.9716,
    "elapsed": 0.2513
  }
]


If no object is detected, empty data is output as follows:


[];


If an error is returned, it is output in the following json format:


{
  "error": {
    "code": 1,
    "message": "Invalid parameters"
  }
}


3.2.3. yaml


Outputs object detection results in yaml format.


- class: car # First object, object class
  area: # Object area (in pixels)
    x: 2171 # Top-left x coordinate
    y: 2281 # Top-left y coordinate
    width: 396 # Width
    height: 521 # Height
  conf: 0.9678 # Object detection confidence (range: 0 ~ 1)
  elapsed: 0.3190 # Elapsed time (seconds)
  licensePlate: # License plate(s)
    - text: 51조8969 # License plate number
      area: # License plate area (in pixels)
        x: 2420 # Top-left x coordinate
        y: 2295 # Top-left y coordinate
        width: 110 # Width
        height: 81 # Height
        angle: 147.5933 # Tilt angle (degrees)
      attrs: # License plate attributes
        ev: true # Eco-friendly electric vehicle status
      conf: # Confidence (range: 0 ~ 1)
        ocr: 0.9005 # OCR confidence
        plate: 0.7864 # Plate recognition confidence
      elapsed: 0.3226 # Elapsed time (seconds)
- class: car # Second object
  area:
    x: 264
    y: 2266
    width: 433
    height: 543
  conf: 0.9706
  elapsed: 0.3191
  licensePlate:
    - text: 41노7656
      area:
        x: 315
        y: 2281
        width: 103
        height: 81
        angle: 211.3135
      conf:
        ocr: 0.916
        plate: 0.8299
      attrs:
        ev: false
      elapsed: 0.5527
- class: car # Third object (when there is no license plate information)
  area:
    x: 777
    y: 0
    width: 579
    height: 403
  conf: 0.9716
  elapsed: 0.3191


If no object is detected, empty data is output as follows:




If an error is returned, it is output in the following yaml format:


error
  code: 1
  message: Invalid parameters


3.2.4. xml


객체 인식 결과를 xml 형식으로 출력합니다.


<?xml version="1.0" encoding="utf-8"?>
<data>
  <!-- First object
    class: object class
    conf: object detection confidence (range: 0 ~ 1)
    elapsed: elapsed time (seconds)
  -->
  <object class="car" conf="0.9679" elapsed="0.3287">
    <!-- Object area (in pixels)
      x: top-left x coordinate
      y: top-left y coordinate
      width: width
      height: height
    -->
    <area x="2171" y="2281" width="396" height="521"/>

    <!-- License plate
      text: license plate number
      elapsed: elapsed time (seconds)
    -->
    <license-plate text="51조8969" elapsed="0.3961">

      <!-- License plate attributes
        ev: eco-friendly electric vehicle status
      -->
      <attrs ev="true"/>

      <!-- License plate area (in pixels)
        x: top-left x coordinate
        y: top-left y coordinate
        width: width
        height: height
        angle: tilt angle (degrees)
      -->
      <area x="2420" y="2295" width="110" height="81" angle="147.5933"/>

      <!-- Confidence (range: 0 ~ 1)
        ocr: OCR confidence
        plate: plate recognition confidence
      -->
      <conf ocr="0.9005" plate="0.7864"/>
    </license-plate>
  </object>

  <!-- Second object -->
  <object class="car" conf="0.9706" elapsed="0.3287">
    <area x="264" y="2266" width="433" height="543"/>
    <license-plate text="41노7656" elapsed="0.4364">
      <attrs ev="false"/>
      <area x="315" y="2281" width="103" height="81" angle="211.3135"/>
      <conf ocr="0.9160" plate="0.8299"/>
    </license-plate>
  </object>

  <!-- Third object (when there is no license plate information) -->
  <object class="car" conf="0.9716" elapsed="0.3287">
    <area x="777" y="0" width="579" height="403"/>
  </object>
</data>


If no object is detected, empty data is output as follows:


<?xml version="1.0" encoding="utf-8"?>
<data />


If an error is returned, it is output in the following xml format:


<?xml version="1.0" encoding="utf-8"?>
<error code="1" message="Invalid parameters" />


4. Error Code Table


The complete list of error codes is as follows:






code
message
Description






1
Invalid parameters
When function call arguments are invalid




2
File not found
When the input image file does not exist




3
Invalid image
When the input image memory does not match the required format




4
Unsupported image format
When the input image is in an unsupported format




100
License expired
When the license has expired




101
Corrupted library
When some library configuration files are missing or corrupted




102
Not initialized
When the engine has not been initialized




103
Too many workers
When the number of library call threads exceeds the limit (maximum 256)




104
Resource exhausted
When no more resources can be allocated




105
License not installed
When the license is not installed (occurs if the free trial license is not installed on Linux)




106
USB dongle I/O error
When reading the USB license dongle fails




107
License required
When a license is required to use the function




108
Unsupported platform
When running in an unsupported environment




200
Unknown
Other undefined errors