#Optical Character Recogntion for Vehicles
An end-end system to extract number plate characters for indian vehicles.
The project has been developed using TensorFlow for License Plate detection from a given car image and uses the Tesseract Engine to recognize the charactes from the detected plate.
- Anaconda 3 (Tool comes with most of the required python packages along with python3 & spyder IDE)
- Tesseract Engine (Must need to be installed)
- The given data had 237 images which were less to trian a robust model.
- Augmentor library was used to augment the dataset using the following augmentations:
- Horizontal flip
- Rotations ( 30 – 180)
- Random distortions
- Skew of two different types
- Size preserving shearing
- Generated the set of 500 images (Cars along with number plate). Then annotated the set of images by drawing the boundary box over the number plates to send it for the training phase as Augmentor library had.
- The Annotation gives the co-ordinates of license plates such as (xmin, ymin, xmax, ymax)
- Then the co-ordinates are saved into a XML file by Augmentor library.
- All the XML files are grouped and the Co-ordinates are saved in CSV file.
- Then the CSV file is converted into TensorFlow record format.
- The set of other separate 10 images also gone through the above steps and saved as Test Record file
- Tensorflow-gpu version used to send the set of annotated images were sent into the YOLOv3 network (Tf object detection API was used), where the metrics such as model learning rate, batch of images sent into the network and evaluation configurations were set. The training For object detection was done on Colab. For OCR: tesseract trained on a dataset of 35000 low quality digits dataset.
- Then the detected number plate is cropped using Tensorflow, By using the Google Tesseract-OCR (Package originally developed to scan hard copy documents to filter out the characters from it) the picture undergoes some coversions using computer vision package then the charcters are filtered out.
- The basic motion capturing has been implemented to capture the picture of moving vehicle by using the openCV library where the sampling rate (i.e. fps) is taken as input for running the application and inference is performed on each frame