Skip to content

Automatic number plate recognition (ANPR) with the particularity that it was trained with 7200+ images of Argentine license plates.

License

Notifications You must be signed in to change notification settings

LeonardoFaggiani/JustAnotherAlpr

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

78 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Just Another ALPR

This is a proof of concept of how to train my own YOLO model to detect vehicle license plates.

Darknet is used as a framework and the YOLO algorithm to detect license plates, YOLOv4-tiny is used because the main idea is to be able to detect license plate in real time.

On the other hand, I used Python to draw and manipulate the steaming of the video.

The model was trained entirely with images of Argentine license plates (7200 images). If you want to obtain the images that were used to train your own model, you can download them here.

Alt Text

Installation & Requirements

  1. Install the Darknet framework, I recommend the following repository https://github.com/hank-ai/darknet
  2. Clone https://github.com/LeonardoFaggiani/JustAnotherAlpr.git
  3. Go to the darknet folder and modify the darknet.py file on line 328 "{path/to/your}/darknet.dll" replace {path/to/your} with the path where the darknet installer is located, save it.
  4. Go to the nn/license-plate path and modify the license-plate.data file and you must tell darknet where your license-plate.names file is, for example if you cloned the JustAnotherAlpr repository on the C drive, it would be C://JustAnotherAlpr/nn/license-plate/license-plate.names

Usage

UI

Go to the root project and execute python -m ui.menu Alt Text

Video/Webcam/Images

Process a Video File

python darknet_video.py --input samples\Testing-Driving.avi --weights ..\nn\license-plate\license-plate.weights --config_file ..\nn\license-plate\license-plate.cfg --data_file ..\nn\license-plate\license-plate.data

Use Display for Real-Time Detection

python darknet_video.py --input 0 --weights ..\nn\license-plate\license-plate.weights --config_file ..\nn\license-plate\license-plate.cfg --data_file ..\nn\license-plate\license-plate.data

Use Webcam for Real-Time Detection

python darknet_video.py --input 1 --weights ..\nn\license-plate\license-plate.weights --config_file ..\nn\license-plate\license-plate.cfg --data_file ..\nn\license-plate\license-plate.data

Save Processed Video

python darknet_video.py --input samples\Testing-Driving.avi --out_filename processed_Testing-Driving.avi --weights ..\nn\license-plate\license-plate.weights --config_file ..\nn\license-plate\license-plate.cfg --data_file ..\nn\license-plate\license-plate.data

Single Image Detection

python darknet_images.py --input samples\Testing-Image.jpg --weights ..\nn\license-plate\license-plate.weights --config_file ..\nn\license-plate\license-plate.cfg --data_file ..\nn\license-plate\license-plate.data --gpu

License Plate Metrics

image

image

TODO

  • Upload differents kind of images to train the model.
  • Create Model
  • Draw boxes in video streaming
  • OCR
  • Information of the owner of the vehicle license (Simulation)
  • Move images to blob storage
  • Create GUI with Tkinter 🛠️ (In Progress)
  • Improve model (motorbike license plate)

About

Automatic number plate recognition (ANPR) with the particularity that it was trained with 7200+ images of Argentine license plates.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Languages