About Me
(in about 309 words)
Intro
I am a data science enthusiast and an aspiring machine learning practitioner with core interest in deep learning. My current interest lies in the fields of Computer Vision, Natural Language Processing and MLOps.
Currently
I am working as a Data Science and Machine Learning Engineer at Neovotech. I am also working on making a dataset and a research paper on Automatic Recognition of Handwritten Bangla texts from images/pdfs.
Notable Works
Automatic detection and recognition of handwritten Bengali texts from images.
- At the start of this project there were no dataset for Bengali word recognition from documents. Hence, I collected 303 documents from different individuals and created a dataset of Handwritten documents for the Bengali language.
- I made an entire pipeline that takes in a document with handwritten paragraph in Bengali, and returns digitized text of the document. The pipeline consists of, image pre-processing, line segmentation, removal of overlapping lines, word detection and finally word recognition. The recognition accuracy was 76% at that time.
Learning without Forgetting for 3D Point Cloud Objects.
- In this paper we mainly experimented learning without forgetting on 3D point clouds and knowledge distillation with semantic vectors to minimize the phenomenon of catastrophic forgetting.
- My main contribution was experimenting with 2D images and testing the 3D models.
A pipepline for automatic OMR and OCR from OMR sheets
- This was a product made for a school where a scanned OMR sheet was provided as an input and the OMR and OCR of the handwritten marks was returned as an output.
- My main contribution was creating the OCR part along with the pipeline, integrating the OMR part, creating custom dataset, image pre-processing, page alignment, testing and setting up connection with mqtt.
I dream of
- getting a PHD in data science
- contributing and making an impact
- helping people through my work
