Arnab Dey

Career ​Portfolio

Data Science
Deep Learning
Computer Vision
Robotics
Web Development

MSc Thesis at Politecnico Di Milano

Human activity and emotion recognition from RGB videos using deep learning (2020) 

Demo                  GitHub                 Paper 
Deep learning based approach to recognize human activities and emotions from RGB videos in real-time for robotics application. 
  • Implemented using PyTorch and TensorFlow.
  • Used NVIDIA Jetson Nano and Google Coral for deployment.
  • Transfer learning techniques were used for training the models

Data Scientist Intern at Unicredit (Sept 2019 - Mar 2020)

OCR pipeline

  • Implementation of a generic and configurable OCR web service for text extraction from scanned documents.
  • Configurable image processing based preprocessing modules, configurable modules for different OCR engines, Language sensitive word correction post processing modules.

Garnishment document classification and enrichment

  • Testing different enrichment models and perform different tests.
  • Implementation of Deep learning techniques for document classification and named-entity recognition.  
  • Dockerizing, Integration and testing web services.

Stamp recognition and information extraction

  • Specific stamp recognition from scanned documents and date and time extraction.
  • Implemented with Keras, scikit-learn, OpenCV.
  • Managed to achieve 83% accuracy
Picture

E-commerce website Lectect​

at Politecnico di Milano (2019)

GitHub        Live Website
  • It’s an online book store.
  • Worked with a team of 3.
  • Implemented with node.js, Bootstrap, CSS, Html
Picture

Art with Computer Vision ​

at Politecnico Di Milano (2019)

GitHub
  • Camera calibration from rectangular paper.
  • 3D reconstruction of the paper and localization of pen on paper.
  • Implemented with Python. 

Bob: Dancing robot (2019)

at Politecnico di Milano

  • Designed a 40cm tall robot capable of dancing with music by moving arms, head and body.
  • The robot able to analyze music using microphone and dance with beats, also it can detect human using sensors.
  • Using Arduino. Worked with a multidisciplinary team.
Picture

Ranky (2018)

Official mobile app for International Olympiad in Informatics

 Google Play        GitHub
  • It’s a cross platform mobile application made with React Native. This application made for IOI competitions.
  • Worked with a team of 2 members​

Dumb & Drunk 

at Politecnico Di Milano (2018)

GitHub                  Game Play
  • It’s a multiplayer PC game implemented with Unity.
  • Worked with a team of 5 members.
  • Worked as 3D artist and programmer on this project.
Picture

Donothrow, a web service for smart city

at Politecnico Di Milano (2018)

GitHub          Live Demo
  • Implemented with nodejs, html, and swagger for API documentation
Picture

B.Tech thesis at WBUT (2017)

Automated region selection in images based on visual saliency

  • Achieved automated region selection without prior knowledge and user interaction.
  • Implemented using Matlab.

Hand gesture recognition with RGB camera

at Indian Statistical Institute, Kolkata (2016)

Demo
  • Human hand gesture recognition and replace mouse action with hand gestures for Linux system.
  • C++, OpenCV, Xlib library for Linux.
  • Achieve real-time performance with RGB camera.
Site powered by Weebly. Managed by Porkbun
  • Home
  • Discover
  • Journal
  • Portfolio
  • Contact
  • Career Portfolio
  • Home
  • Discover
  • Journal
  • Portfolio
  • Contact
  • Career Portfolio