https://github.com/yycorcino/license-plate-recognition

CNN model to classify characters on Japanese license plates
https://github.com/yycorcino/license-plate-recognition

cnn cscea485 python tensorflow

Last synced: 27 days ago
JSON representation

CNN model to classify characters on Japanese license plates

• Host: GitHub
• URL: https://github.com/yycorcino/license-plate-recognition
• Owner: yycorcino
• License: other
• Created: 2024-12-11T18:29:12.000Z (about 1 month ago)
• Default Branch: main
• Last Pushed: 2024-12-11T18:36:24.000Z (about 1 month ago)
• Last Synced: 2024-12-11T19:41:47.021Z (about 1 month ago)
• Topics: cnn, cscea485, python, tensorflow
• Language: Jupyter Notebook
• Homepage:
• Size: 2.16 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: license-plate-recognition.ipynb

Awesome Lists containing this project

README

        




  
License Plate Recognition



### About The Project

The goal of the project is to develop a license plate recognition system to identify and capture Japanese vehicle plate numbers.

#### Use Cases

##### Crime Prevention

The rise in vehicles on the road makes identifying criminal activity increasingly challenging. License plate recognition assists law enforcement by enabling quick identification of stolen or suspicious vehicles. 

##### Smart Garage Opener

License plate recognition can be used to automate opening the garage door when it detects your car is heading toward the garage, essentially a virtual key to your garage.

#### Approach

The proposed approach is using a Convolutional Neural Network model (CNN). CNN models excel in processing spatial data, such as images, especially when much noise or additional entities obscure the key features. In the dataset, many of images consists of features (text) shifted in different angles, different lighting, and different image quality. CNN is a great option to extract these features by training on the dataset through many of its layers.

#### Important Layers

##### Layer 2 - First Convolutional Layer (32 filter)

The purpose of this layer is to detect features from edges and shapes.

##### Layer 3 - First Max Pooling Layer

The purpose of this layer is to reduce the spatial size which helps the model focus closer to the features (text).

##### Layer 4 - Second Convolutional Layer (64 filter)

The purpose of this layer is to allow the model to better detect complex patterns.

##### Layer 5 - Second Max Pooling Layer

The purpose of this layer is to again reduce spatially and the main benefit of this is to help detect the text invariance.

### Dataset 

The data set is sourced from [Kaggle](https://www.kaggle.com/datasets/nickyazdani/license-plate-text-recognition-dataset).

Characteristics of the dataset:

- cropped images of Japanese license plates

- license plate photos are captured in various angles and lighting

- license plate exist various styles and color

**Note** - Notebook consists of borrowed code from [published CNN Model](https://keras.io/examples/vision/captcha_ocr/) and adjusted to fit application.