Number Plate Recognition with TensorFlow

This repository is going to implement a simple number plate recognition using CNN model with TensorFlow. Please notice that this program can only recognize the pattern of the number plate as follow:

• More information about number plate recognition with Tensorflow

---

File Structure

NumberPlate_Recognition     # This is ./ in this repository
|--- input/                 # Input image of number plate
|--- output/                # Output image of number plate's recognition
|--- common.py
|--- detect.py
|--- model.py

---

Prerequisite

• Before executing, you need to install the following packages
  • Install OpenCV using pip

    $ [sudo] pip install opencv-python

  • Install TensorFlow using pip

    $ [sudo] pip install -U tensorflow

• When installing TensorFlow, you may meet the following problems. However, if install successfully, you can follow the execution
  • Error message: launchpadlib 1.10.3 requires testresources, which is not installed.

    $ [sudo] pip install launchpadlib

  • Error message: Cannot uninstall 'enum34'. ...

    $ sudo apt-get remove python-enum34

---

Execution

NOTICE: The image of number plate should be placed in the folder ./input/

• Execution

  # Make sure your current directory is "src/"
  $ python detect.py t1.jpg weights.npz
  $ python detect.py t2.jpg weights.npz

• Open the folder ./out/ to see the result of recognition. The filename of the result will same as the input file.

  $ ls out/
  t1.jpg  t2.jpg

---

References

• TensorFlow
• opencv-python 3.4.1.15
• THE MNIST DATABASE

---

Contributor

NOTICE: You can follow the contributing process CONTRIBUTING.md to join me. I am very welcome any issue!

• David Lu

---

License

GNU GENERAL PUBLIC LICENSE Version 3