cr-plates-generator

Costa Rica license plate generator for computer vision

Features

• Plate formats: car, motorcycle, trucks, taxi, disabled drivers.
• Perspective rotations, size and random backgrounds.
• Annotations with character bounding boxes (class, cx, cy, w, h)
• configuration.cfg and templates.json files for customization.

Examples

[Pending]

Getting Started

Prerequisites

• Python >= 3.6.0
• Modules on requirements.txt pip3 install -r requirements.txt

Running sample code

1. Add some images to ./backgrounds directory, these will work as random backgrounds for the plates. SUN Database is recommended.

2. Run the main file python ./main.py

3. Random plates will be generated on ./output directory

Settings

configuration.cfg

The following is a description of all the settings on this file.

Setting	Description	Value
[General]
dataset_size	Quantity of images to generate	int
templates_path	Path to directory containing base plate images	string
templates_config	Path to JSON configuration for each type of plate	string
[Image]
resize_plate	Apply resizing to the base plate images setting	bool
plate_scales	List of scaling factors to be used	list
resize_bg	Apply resizing to background images to a fixed size	bool
bg_sizes	List of target (width,height) pairs to resize bgs	list
draw_bboxes	Whether to draw bounding boxes (Use for testing only)	bool
bbox_padding	Spacing between bbox and inner object (px)	int
[Perspective]
theta_range	Maximum angle (degrees) to rotate plate over z-plane	float
phi_range	Maximum angle (degrees) to rotate plate over y-plane	float
gamma_range	Maximum angle (degrees) to rotate plate over x-plane	float

templates.cfg

[Pending]

Authors

• Alejandro Pereira - Home Page

License

This project is licensed under the GNU General Public License v3.0 - see the LICENSE.md file for details

References

• Perspective transformations: https://nbviewer.jupyter.org/github/manisoftwartist/perspectiveproj/blob/master/perspective.ipynb
• Another plate generator: https://matthewearl.github.io/2016/05/06/cnn-anpr/
• SUN Database: https://vision.princeton.edu/projects/2010/SUN/