I developed an end-to-end handwritten optical character recognition pipeline on the IAM dataset, combining computer vision and deep learning. For my M.Tech dissertation, I segmented characters with Otsu thresholding and contour sorting, then fine-tuned a ResNet-18 with CBAM attention to achieve 75.40% validation accuracy.
In my Java security project, I developed a desktop application to encrypt, hide, and restore local files using AES, with email-based OTP authentication. I also trained a flower classifier that reached 90% average test accuracy and built a Tkinter interface for image predictions.
