Moseli Mots'oehli

DeepsMoseli.github.io

moselim@hawaii.edu

DeepsMoseli

moselim@hawaii.edu

1680 East-West Rd, Post 314-6

Honolulu, Hawaii 96822 USA

808-476-2232

Interests

Deep learning (Autonomous Perception, Weakly Supervised Learning, Deep Active Learning, Label Noise, Computer Vision), Software engineering (react.js, Meteor, FastAPI, TypeScript), Start ups

Skills

Languages and Frameworks: Python, Pytorch, Tensorflow, React.js, TypeScript, C++, Meteor

Toolsets/OS: Windows, Linux, AWS, Git/GitHub

Education

University of Hawaii, Honolulu, HI

Ph.D., Computer Science

  • Autonomous Perception, Deep Active Learning, Label Noise

2021 - Present

University of Hawaii, Honolulu, HI

M.S, Computer Science

  • Deep Learning, Computer Vision, Software Engineering II

2019 - 2021

University of Pretoria, South Africa

M.S., Data science

  • Generative Adversarial Networks, Natural Language Processing, Machine Learning

2017 - 2019

Work

Data Science Lead, The Shard - South Africa

www.theshard.co.za/

Technical client engagements, Project Scoping, Research, Architecture Design, Overseeing Implementation

  • African bank, Capitec bank, National treasury

2022 - 2024

Graduate Research Fellow, Hawaii Data Science Institute - University of Hawaii

datascience.hawaii.edu/students/fellows/

Using Computer vision and Deep learning methods to help improve marine life preservation efforts.

2020 - 2022

Summer Research Assistant, University of Hawaii at Manoa, ICS

Using Computer vision and Deep learning methods to help improve marine life preservation efforts.

2020 - 2020

Teaching Assistant, University of Hawaii at Manoa, ICS

I assist students during my office hours, set up assignments, recitations, grade assignments and projects

  • Computer Vision, Intro to Data Science, Machine Learning, Algorithms

2019 - Curr

Credit risk data analyst, Vodacom South Africa

www.vodacom.co.za/

  • Building and maintaining predictive models
  • Automation of credit tracking processes using R and SAS

2015 - 2017

Publications

Balancing Accuracy and Efficiency in Vision Transformers for Noisy Annotations: Insights on Patch and Embedding Size, Mots'oehli .M, Gong .J.Y, Mogale .H, (To be submitted), 2024

#

GCI-ViTAL: Gradual Confidence Improvement with Vision Transformers for Active Learning on Label Noise, Mots'oehli .M, Baek .K, (Under review), 2024

https://arxiv.org/abs/2411.05939

Assistive Image Annotation Systems with Deep Learning and Natural Language Capabilities: A Review, Mots'oehli .M, IEEE International Conference On Emerging Trends In Networks And Computer Communications (ETNCC 2024), 2024

https://arxiv.org/abs/2407.00252

FishNet: Deep Neural Networks for Low-cost Fish Stock Estimation, Mots'oehli .M, Nikolaev .A, IGede .W.B, Lynham .J, Mous .P.J, Sadowski .P, IEEE International Conference on Omni-Layer Intelligent Systems (COINS 2024), 2024

https://arxiv.org/abs/2403.10916

Comparision of Adversarial and Non-adversarial LSTM Music Generative Models, Mots'oehli .M, Bosman. A, De Villiers .JP, Science and Information Conference (SAI 2023) (Springer), page 428, 2023

https://link.springer.com/chapter/10.1007/978-3-031-37717-4_28

PuoBERTa: Training and evaluation of a curated language model for Setswana, Marivate .V, Mots'oehli .M, Wagnerinst .V, Lastrucci .R, Dzingirai .I, Southern African Conference for Artificial Intelligence Research (SACAIR 2023) (Springer), pages 253–266 , 2023

https://arxiv.org/abs/2310.09141

Deep Active Learning in the Presence of Label Noise: A Survey, Mots'oehli .M, Baek. K, Ph.D. Portfolio Literature Review, 2023

https://arxiv.org/pdf/2302.11075.pdf

Public Parking Spot Detection and Geo-localization Using Transfer Learning, Mots'oehli .M, Yang .Y, Proceedings of the 3rd Southern African Conference for Artificial Intelligence Research (SACAIR 2022), page 109, 2022

https://arxiv.org/abs/2209.00213

References

Available upon request