ALPR-Yolov5 (Automatic license Plate detection and recognition)

This is the Automatic license plate detection and recognition system using Yolov5. Both plate detection and character detection and recognition using Yolov5. I used EnglishLP dataset for experiment but you can try with any other dataset also

Alt text

dependencies

NVIDIA GPUs
NVIDIA graphic driver
CUDA toolkit
cuDNN library
OpenCV 
Cython
matplotlib>=3.2.2
numpy>=1.18.5
opencv-python>=4.1.2
pillow
PyYAML>=5.3
scipy>=1.4.1
tensorboard>=2.2
torch==1.6.0
torchvision==0.7.0
tqdm>=4.41.0

This project is tested under following conditions:

Ubuntu 20.04 LTS
NVIDIA GeForce RTX 2080 Super
NVIDIA graphic driver version: 440.95.01
CUDA version: 10.2
cuDNN version: 7.6.5
OpenCV version: 4.2

Installations

Install NVIDIA graphic driver, CUDA toolkit, cuBLAS, and cuDNN

Configure conda virtual environment

Plate detection training for custom custom data

This guide explains how to train your own custom dataset with YOLOv5 for plate detection

Before You Start

Clone this repo, prepared datasetin yolo format( you can check inside voc folder for sample image and label), and install dependencies, including Python>=3.8 and PyTorch>=1.7.

$ git clone https://github.com/bharatsubedi/ALPR-Yolov5  # clone repo

Train On Custom Data

update data/voc.yaml data/voc.yaml, shown below, is the dataset configuration file that defines 1) a path to a directory of training images (or path to a *.txt file with a list of training images), 2) the same for our validation images, 3) the number of classes, 4) a list of class names:
```
train: voc/images/
val: voc/images/
```

number of classes

nc: 1

class names

names: ['plate']

- create labels: 
   After using a tool labelImg to label your images, export your labels to YOLO format, with one *.txt file per image (if no objects in image, no *.txt file is required). The *.txt file specifications are:

    * One row per object
    * Each row is `class x_center y_center width height` format.
    * Box coordinates must be in normalized xywh format (from 0 - 1). If your boxes are in pixels, divide `x_center` and `width` by image width, and `y_center` and `height` by image height.
    * Class numbers are zero-indexed (start from 0).
- Organize directories
Organize your train and val images and labels according to the example below. In this example we assume /voc is next to the /yolov5 directory. YOLOv5 locates labels automatically for each image by replacing the last instance of /images/ in the images directory with /labels/. For example:

voc/images/000000109622.jpg # image voc/labels/000000109622.txt # label

For further understanding you can check inside `plate_detection` or `charater_detection`
### Select a Model
Select and modify the models parameters `nc` related to your dataset

plate_detection/yolov5/models/(yolov5s,yolov5m,yolov5l,yolov5x) character_detection/yolov5/models/(yolov5s,yolov5m,yolov5l,yolov5x)

### Train
you can modify hyperparameter inside train.py

Train YOLOv5s on COCO128 for 5 epochs $ python train.py --img 640 --batch 16 --epochs 5 --data data/voc.yaml


### End to end testing
copy plate detection weight and character detection weight into `end-t-end-test/final_weight/(detection_weight).pt and end-t-end-test/recognition_model/final_weight/(character recognition weight).pt`
For more information check each folder inside 

EnglishLP dataset pretrained weight download link for 
* detection weight https://drive.google.com/file/d/1dnTTVbGq4NLDDlJHMZBAwPcRVjL_yQgZ/view?usp=sharing
* recognition weight https://drive.google.com/file/d/1aNvYucWlfVy8w_Ijh0wEE9j7rtMCKLjR/view?usp=sharing

bharatsubedi / ALPR-Yolov5

readme