license-plate-recognition-barrier-0007

Use Case and High-Level Description

This model uses a small-footprint network trained end-to-end to recognize Chinese license plates in traffic.

Validation Dataset - Internal

300320 Sythetic Chinese plates i.e. the plate text on them consists of symbols generated randomly (but to conform to the plate requirements in terms of the number of characters, sequence, shape, placement, etc.). The "real-looking" appearance of the plates (rotation, dirt, color, lighting, etc.) is achieved by a style transfer procedure.

Example

Note: The license plates on the image were modified to protect the owners' privacy.

Specification

Metric Value
Rotation in-plane ±10˚
Rotation out-of-plane Yaw: ±45˚ / Pitch: ±45˚
Min plate width 94 pixels
Ratio of correct reads 98%
GFlops 0.347
MParams 1.435
Source framework TensorFlow*

Limitations

Only "blue" license plates, which are common in public, were tested thoroughly. Other types of license plates may underperform.

Inputs

Original Model

Image, name: input, shape: 1, 3, 24, 94, format is 1, C, H, W, where:

  • C - channel
  • H - height
  • W - width

Channel order is BGR.

Converted Model

Image, name: input, shape: 1, 3, 24, 94, format is 1, C, H, W, where:

  • C - channel
  • H - height
  • W - width

Channel order is BGR.

Outputs

Original Model

Encoded vector of floats, name: d_predictions, shape: 1, 88, 1, 1. Each float is an integer number encoding a character according to this dictionary:

    0 0
    1 1
    2 2
    3 3
    4 4
    5 5
    6 6
    7 7
    8 8
    9 9
    10 <Anhui>
    11 <Beijing>
    12 <Chongqing>
    13 <Fujian>
    14 <Gansu>
    15 <Guangdong>
    16 <Guangxi>
    17 <Guizhou>
    18 <Hainan>
    19 <Hebei>
    20 <Heilongjiang>
    21 <Henan>
    22 <HongKong>
    23 <Hubei>
    24 <Hunan>
    25 <InnerMongolia>
    26 <Jiangsu>
    27 <Jiangxi>
    28 <Jilin>
    29 <Liaoning>
    30 <Macau>
    31 <Ningxia>
    32 <Qinghai>
    33 <Shaanxi>
    34 <Shandong>
    35 <Shanghai>
    36 <Shanxi>
    37 <Sichuan>
    38 <Tianjin>
    39 <Tibet>
    40 <Xinjiang>
    41 <Yunnan>
    42 <Zhejiang>
    43 <police>
    44 A
    45 B
    46 C
    47 D
    48 E
    49 F
    50 G
    51 H
    52 I
    53 J
    54 K
    55 L
    56 M
    57 N
    58 O
    59 P
    60 Q
    61 R
    62 S
    63 T
    64 U
    65 V
    66 W
    67 X
    68 Y
    69 Z

Converted Model

Encoded vector of floats, name: d_predictions:0, shape: 1, 88. Each float is an integer number encoding a character according to this dictionary:

    0 0
    1 1
    2 2
    3 3
    4 4
    5 5
    6 6
    7 7
    8 8
    9 9
    10 <Anhui>
    11 <Beijing>
    12 <Chongqing>
    13 <Fujian>
    14 <Gansu>
    15 <Guangdong>
    16 <Guangxi>
    17 <Guizhou>
    18 <Hainan>
    19 <Hebei>
    20 <Heilongjiang>
    21 <Henan>
    22 <HongKong>
    23 <Hubei>
    24 <Hunan>
    25 <InnerMongolia>
    26 <Jiangsu>
    27 <Jiangxi>
    28 <Jilin>
    29 <Liaoning>
    30 <Macau>
    31 <Ningxia>
    32 <Qinghai>
    33 <Shaanxi>
    34 <Shandong>
    35 <Shanghai>
    36 <Shanxi>
    37 <Sichuan>
    38 <Tianjin>
    39 <Tibet>
    40 <Xinjiang>
    41 <Yunnan>
    42 <Zhejiang>
    43 <police>
    44 A
    45 B
    46 C
    47 D
    48 E
    49 F
    50 G
    51 H
    52 I
    53 J
    54 K
    55 L
    56 M
    57 N
    58 O
    59 P
    60 Q
    61 R
    62 S
    63 T
    64 U
    65 V
    66 W
    67 X
    68 Y
    69 Z

Download a Model and Convert it into Inference Engine Format

You can download models and if necessary convert them into Inference Engine format using the Model Downloader and other automation tools as shown in the examples below.

An example of using the Model Downloader:

python3 <omz_dir>/tools/downloader/downloader.py --name <model_name>

An example of using the Model Converter:

python3 <omz_dir>/tools/downloader/converter.py --name <model_name>

Legal Information

The original model is distributed under the Apache License, Version 2.0. A copy of the license is provided in APACHE-2.0.

[*] Other names and brands may be claimed as the property of others.