redwankarimsony Goto Github PK

Name: Redwan Sony

Type: User

Company: Michigan State University

Bio: Ph.D. Student @iPRoBe-lab, Michigan State University. Former Lecturer at IUT, Bangladesh. Explainability, Interpretability, Forensic Biometrics.

Twitter: RedwanKarimSony

Location: East Lansing, Michigan, USA

Hello, This is Redwan Karim Sony!
This contains overview of the my activities like research works, machine learning competitions, projects etc. Notworthy works are listed here with necessary links. Star them if you see them worthy of..

Research & Publications:

No.	Description	Published	GitHub Repo	Link to Publication
1.	Advanced Agglomerative Clustering Technique for Phylogenetic Classficiation using Manhattan Distance.	Yes	Code	BIOCOMP'17
2.	Performance Comparison of Feature Descriptors in Offline Signature Verification	Yes	Code	JET Link
3.	Less is More: Lighter and Faster Deep Neural Architecture for Tomato Leaf Disease Classification	Yes	Code	IEEE Access
4.	Investigating Weight-Perturbed Deep Neural Networks With Application in Iris Presentation Attack Detection	Yes	Code	WACV Workshop 2024

iPRoBe Lab Projects:

No.	Description	Published	GitHub Repo	Link to Publication
1.	Comparative Medical Image Radiography for Identification Task.	No	Code	~

ML Competitions on HackerEarth

The following table contains all the code bases of the competitions that I participated on HackerEarth. The original repository is uploaded in my GitHub account in their respective repos. However, for easier browse through specific problem and solutions, this table may come handy to you. Keep Learning!

No.	Challenge Name	GitHub Repo	Type	Position	LeaderBoard
1.	HackerEarth Deep Learning Challenge: Snakes in the hood	Solution*	Classification	3rd among 3389 participants (Top 1%)	Link
2.	HackerEarth Machine Learning Challenge: Are your employees burning out?	Solution	Regression	40th among 560 participants (Top 7%)	Link
3.	HackerEarth Machine Learning Challenge: Carnival Wars	Solution*	Regression	310 th out of 2240 teams (top 13%)	Link

* Note: If you can't find the GitHub Repo of the solution, it means that the competition is live, hence the repo is private for now. Once the contest is over, I will make that repo public.

ML Competitions on Kaggle

The following table contains all the code bases of the competitions that I participated on Kaggle. Most of the solution are done through kaggle kernel. In order to view the live original notebook with dataset loaded, follow the kaggle live link for respective notebooks in the github repos. However, a copy of them are uploaded in their respective github repos. For easier browse through specific problem and solutions, this table may come handy to you. Keep Learning!

No.	Challenge Name	GitHub Repo	Type	Position	LeaderBoard
1.	OSIC Pulmonary Fibrosis Progression	Solution	Regression	Solo Bronze medal 198th out of 2097 teams (Top 10%)	Link
2.	SIIM-ISIC Melanoma Classification	Solution	Classification	Solo 386th out of 3314 teams (Top 12%)	Link
3.	Flower Classification with TPUs	Solution	Classification	Solo 84th out of 848 teams (Top 10%)	Link
4.	Plant Pathology 2020 - FGVC7	Solution	Classification	103rd out of 1317 teams (Top 8%)	Link
5.	RSNA-STR Pulmonary Embolism Detection	Solution	Multistage Classification	Solo 288th out of 784 teams (Top 37%)	Link
6.	Jigsaw Multilingual Toxic Comment Classification	Solution	Classification, NLP	Solo 330th out of 1621 teams (Top 21%)	Link
7.	Sorghum -100 Cultivar Identification - FGVC 9	Solution	Classification	SOlo 43 th out of 252 teams (top 17%)	Link

Miscellaneous Competitions

No.	Challenge Name	GitHub Repo	Type	Position	LeaderBoard
1.	Dhaka-Ai-Traffic-Detection-Challenge	Repo Link	Object Dectection and Localization	Live Contest*	Look for team name init()

Projects and Tutorial Repos

No.	GitHub Repo	Description	Category
1.	Object-Detection-and-Segmentation-with-TorchVision	For this tutorial, I have finetuned a pre-trained Mask R-CNN model in the Penn-Fudan Database for Pedestrian Detection and Segmentation. It contains 170 images with 345 instances of pedestrians, and later on it will use it to illustrate how to use the new features in torchvision in order to train an instance segmentation model on a custom dataset.	Detection and Segmentation

Author:

Md. Redwan Karim Sony
Lecturer,
Department of Computer Science & Engineering (CSE),
Islamic University of Technology (IUT)
Dhaka, Bangladesh
Google Scholar | ResearchGate | Kaggle | Linkedin |Academic CV
Email: [email protected]
[email protected]

Recommend Projects