University of South Dakota AI Research Lab (formerly 2ai lab)
Mission: The Coyote's AI powerhouse on Sustainable innovation from the heart of Rushmore State
Website: ai-research-lab.org •
Contact: [email protected] •
Location: Vermillion, SD, USA
We advance foundational AI and machine learning with a focus on sustainable, green computing to ensure efficiency and minimize carbon impact. Our interdisciplinary research spans computer vision, data mining, pattern recognition, and big data, impacting fields like healthcare, biometrics, forensics, speech, and IoT.
Join us as we drive AI innovation with sustainability at its core!
MFil-Mamba: Multi-Filter Scanning for Spatial Redundancy-Aware Visual State Space Models
Multi-filter scanning based visual state-space architecture that captures diverse spatial dependencies while reducing redundancy in image representations.
Winsor-CAM
Class activation mapping technique using winsorization for more stable and interpretable visual explanations.
AI-CARE: Carbon-Aware Reporting Evaluation Metric for AI Models
Framework for evaluating AI models with respect to performance, energy usage, and carbon impact.
I Detect What I Don't Know: Incremental Anomaly Learning with SWAG for Oracle-Free Medical Imaging
Unsupervised anomaly detection framework that incrementally learns normal patterns using uncertainty-guided updates without labeled anomalies.
DeepWhaleNet: Climate Change-aware FFT-based Deep Neural Network for Passive Acoustic Monitoring
FFT-based DNN for passive acoustic whale-call detection with climate-aware considerations; built for UPAM workflows.
Non-Uniform Illumination Attack for Fooling Convolutional Neural Networks
NUI masks degrade CNNs; simple defense via NUI-augmented training across CIFAR-10, TinyImageNet, Caltech-256.
SegFast-V2: Semantic image segmentation with fewer parameters
Compact encoder-decoder with kernel factorization & depthwise deconvs; CPU-friendly yet competitive.