Projects


Road Track(s) Identification Using Image Segmentation

The project aims to create a low-latency reliable tool using computer vision technologies to aid with road track identification through image segmentation tasks. A model is developed based on the popular U-Net architecture to achieve this goal.
This work is purely for educative purposes on the potential use of AI in this field/problem space.

Read more about it here.

Research Paper Search APP.

By harnessing the power of a Sentence Transformer model, the system enables context-aware searches, surpassing traditional keyword-based methods.
Researchers can now effortlessly input natural language queries, retrieving more precise and relevant abstracts. This project showcases the potential of advanced NLP techniques in revolutionizing information discovery across various domains, promising a transformative solutions for academia and industry alike.
Try it out here.

Swahili Language Model (MLM Task)

Pre-trained on a large corpus of Swahili data in a self-supervised fashion. Focuses on the Masked Language Modeling (MLM) task. i.e. taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words.

Read more about it here.

Gists

Fine Tuning BERT Uncased for Swahili text classification

    A notebook on fine tuning BERT Uncased MLM model for multi-class text classification of Swahili dataset.

N-Gram Language Modelling

    A brief illustration of language modelling using NLTK Datasets and Library.

Pillow Image Verification

    Was having issues with the image verification for an image classifier. Needed verification to ensure that the image is a real image thus reduce chances of false positives. For this specific task images hade to have some green pigment. This is not ideal but it works for the purpose of this project. Resizing the image to 256 by 256 pixels reduces the area to check thus kind of efficient. If you have a better a way of doing this, please let me know. oluoch9@gmail.com