[NeurIPS 2025] This repo is official PyTorch implementation of the paper "Learning Dense Hand Contact Estimation from Imbalanced Data".

Code for paper "Semantic Diversity-aware Prototype-based Learning for Unbiased Scene Graph Generation (ECCV 2024)"

pytorch implementation of Shrinkage loss in our ECCV paper 2018: Deep regression tracking with shrinkage loss

Deep Regression Tracking with Shrinkage Loss (ECCV 2018).

ECG Arrhythmia Detection with ResNet and Transfer Learning

Shortcut to Nowhere: Demystifying Deep Spurious Regression

The Mulan Framework with Multi-Label Resampling Algorithms

compare the performance of cross entropy, focal loss, and dice loss in solving the problem of data imbalance

software vulnerability detection

Demonstrate the application of machine learning on a real-world predictive maintenance dataset, using measurements from actual industrial equipment.

Submission for HR Analytics Hackathon - AnalysticsVidya.

Applied undersampling and oversampling using SMOTE.

This repository features a machine learning project utilizing the Pima Indians Diabetes Dataset to predict diabetes risk. It explores data preprocessing, model training, and evaluation using techniques such as Naive Bayes and K-Nearest Neighbors (KNN) . The aim is to highlight the impact of various health factors on diabetes prediction.

A machine learning solution for churn prediction using CatBoost, achieving a 0.8464 AUC-ROC through feature engineering and hyperparameter optimization.

Real-time ML data drift detection system monitors model health using KS statistical testing and alerts when retraining is needed.

dau is a Python package that implements Density-Aware Undersampling (DAU), a novel undersampling technique for handling imbalanced datasets.

Real-time fraud detection microservice built with XGBoost, FastAPI, and Docker, implementing an end-to-end machine learning pipeline from EDA to production deployment.

Preprocessing the heart disease dataset: A practical guide to EDA, feature encoding, discretization, and handling class imbalance with SMOTE.

The final project for the CE888: Data Science and Decision Making module (Spring Term) at the University of Essex

Exploration and optimization of a ML pipeline, delving into various techniques for enhancing different stages of ML workflows, including data preprocessing, feature engineering, model selection, and hyperparameter tuning.