NumberPlateCNN

In this project a CNN is trained to recognize number plate from an image. At first a model is trained to recognize number digits and numbers. Then the charector segmentation is performed those segmented charectors are recognized.

Step performed:

1. Trained a CNN model in using Pytorch.

90% training accuracy and 88% test accuracy was obtained after 8 epochs.

Confusion Matrix for test dataset.

2) Loaded the Image and performed charector segmentation.

Original Gray scaled image.

Charectors After segmentation

3) Recognize the charectors after segmantation.

Final prediction result.

Dataset:

The dataset is available in this repository.

Citation for the dataset:

@inproceedings{pant2015automatic,
  title={Automatic Nepali Number Plate Recognition with Support Vector Machines},
  author={Pant, Ashok Kumar and Gyawali, Prashnna Kumar and Acharya, Shailesh},
  booktitle={Proceedings of the 9th International Conference on Software, Knowledge, Information Management and Applications (SKIMA)},
  pages={92--99},
  year={2015}
}