This paper presents a novel approach for Indian Number Plate Recognition using the Convolution Neural Network (CNN). We generated and augmented synthetic data in Indian standards with variant background and font. We also collected and labelled real data of Indian License plates to create a genuine Indian oriented dataset. We compared existing models like SVM, KNN and CRNN in terms of accuracy and inference time, and selected LPRNet. We first trained the existing LPRNet model, which is designed for Chinese License plates on synthetic data and secured 85% accuracy. We then modified the LPRNet model and secured 93% accuracy on the same synthetic data. We applied the post-processing technique of pattern recognition rules for Indian standards of License plates to further improve accuracy. The Proposed system achieved 95% accuracy after fine-tuning with real data. Our model is very lightweight and can easily be deployed on Nvidia-Jetson nano, Intel Neural Compute Stick.

No abstract

Identification of input fields that appear on a document is a crucial requirement while digitizing any document. This paper presents a Deep Learning based approach to detect input fields from a form or document which consists of text, images and input fields like textbox, checkbox. The forms have been crawled and labelled manually to generate a dataset for training Deep Learning models. The YOLO V3 model is trained on the labelled dataset having four classes (static text, static image, input text, checkbox) with 1500 instances. We used bounding box techniques to label the dataset. The paper presents detection of limited types of input fields generally appearing on printed forms. We also discussed how such detection models can scale and sustain higher loads. If given the labelled dataset for other types of input fields, the existing YOLO V3 can be trained for them as well. The model is trained for 3500 iterations and the accuracy achieved is 71 percent.

No abstract

No abstract