Avatar

Rafid Abyaad

Masters Student

Technical University of Munich

Biography

I am a masters student at TUM, focusing on Machine Learning, Robotics and Formal Methods. I have experience as a software engineer from my time with Samsung Research, Bangladesh. My research interests lie in designing intelligent deep models and applying them in various paradigms. I have a keen interest in using deep learning to improve reinforcement learning systems.

Interests

  • Deep Learning
  • Reinforcement Learning
  • Robotics
  • Natural Language Processing

Education

  • Master of Science in Informatics, 2022

    Technical University of Munich

  • B.Sc. Engineering in Computer Science and Engineering, 2017

    Islamic University of Technology

Skills

Pytorch
Chromium

Development frameworks

C/C++
Python

Languages

Linux

Shell

Experience

 
 
 
 
 

Software Engineer

Samsung R&D Institute

Feb 2018 – May 2021 Dhaka
Contributions :

  • Designing and developing Intelligent, user-oriented systems.
  • Working as a Framework Engineer.
  • Contributed to multiple commercialized projects.
  • Open source contributor to the Chromium projects.
 
 
 
 
 

Software Engineering Intern

Samsung R&D Institute

Oct 2016 – Jan 2017 Dhaka
Contributions :

  • Developing a Unix host from scratch using socket programming.
  • Developing a GNOME-like terminal for the Unix Host.

Publications

Rafid Abyaad, Rayhan Kabir, Sakib Hasan (2020). A Novel Approach to Categorize News Articles From Headlines and Short Text. 2020 IEEE Region 10 Symposium (TENSYMP).

PDF

Recent Posts

Apr 25, 2021 17 min read

PnPNet: End-to-End Perception and Prediction With Tracking in the Loop

In this post, we will be looking at the paper PnPNet: End-to-End Perception and Prediction with Tracking in the Loop, by Liang et al., which was published in CVPR 2020 [1]. After defining our task and discussing some related research in this field, we will be looking at the methodology of the paper. Then we will analyze the quantitative results and have a look at the qualitative results. Finally, we will finish it off with some remarks and possible ideas for extension.

Projects

*
All Deep Learning Reinforcement Learning Computer Vision Natural Language Processing Web

Advantage Actor Critic

Implementation of Advantage Actor Critic to solve OpenAI LunarLander environment

Deep Q Learning

Implementation of Deep Q Learning to solve OpenAI LunarLander environment

Double Deep Q Learning

Implementation of Double Deep Q Learning to solve OpenAI LunarLander environment

Dueling Deep Q Learning

Implementation of Dueling Deep Q Learning to solve OpenAI LunarLander environment

Dueling Double Deep Q Learning

Implementation of Dueling Double Deep Q Learning to solve OpenAI LunarLander environment

Policy Gradient (REINFORCE)

Implementation of REINFORCE to solve OpenAI LunarLander environment

Soft Actor Critic

Implementation of SAC to solve OpenAI LunarLander environment

Value Actor Critic

Implementation of Value Actor Critic to solve OpenAI LunarLander environment

Aerial Cactus Identification

Approaches and code for identifying cacti from satellite images

APTOS Blindness Detection

Description of approaches for the APTOS medical image classification Competition

Computer Vision 101

Classifying fruits using computer vision models

Facial Expression Recognition

Vision model to classify facial expression

Quora Insincere Questions Classification

Approaches and codes for short-text binary classification

Recursion Cellular Image Classification

Recursion Cellular Image Classification

SIIM-ACR Pneumothorax Segmentation

SIIM-ACR Pneumothorax Segmentation

Tabular vs Vision Models

Performance analysis of Tabular vs Vision models on Digits-MNIST dataset