Projects

Bayesian Network Structure Adaptation for Continual Learning

• Developed a continual structure adaptation framework that integrates beta-Bernoulli processes for structure inference within the sequential Bayes framework, enabling dynamic evolution of both network depth and width in continual learning scenarios.
• Paper published in ICML 2024.

Human Trafficking Recognition from Sex Worker Ads and Inter-Ad Matching, Fusemachines

• Served as Team Lead, overseeing the design of the machine learning pipeline and setting up the data annotation process.
• Implemented models for image-based and text-based trafficking recognition, contrastive loss-based image search, face-based person identification, and NER-based social handle extraction for inter-ad matching.

Analysis of Radio Panelists Data, Fusemachines

• Worked on analysis of the impact of song quality, commercial length, and time of day on panelists, employing custom metrics and statistical tests to assess song quality.
• Fixed existing data pipeline to solve date inconsistency bugs and added feature engineering pipeline to augment existing dataset.

Waste Type Detection, Fusemachines

• Developed an object detection system by integrating single shot detection and MobileNet architecture with focal loss to determine the type of waste and the intent of the person disposing of it.

Session-based Network Intrusion Detection System, Fusemachines

• Performed a feasibility study on utilizing AutoEncoder-based semi-supervised learning for network anomaly detection, leveraging session data extracted from pcap files.

Nepalese License Plate Recognition, Final Year Undergraduate Project

• Developed a license plate recognition system, tailored for Nepalese license plates, with three key stages: vehicle detection, license plate localization, and Nepalese character recognition.
• Created a license plate localization dataset by annotating Nepalese vehicle images, alongside building a Nepalese character recognition dataset using Devanagari fonts.