Resource-Constrained AI Models for On-Device Video Capsule Endoscopy
Bachelor’s Thesis / Master’s Thesis / Student Research Project
Abstract
The Video Capsule Endoscopy is a minimally invasive procedure used to examine the gastrointestinal tract for various pathologies. Unlike many other applications, deploying large neural networks with millions of parameters is often impractical in this context due to the limited computational resources of small medical edge devices. Therefore, our research group focuses on developing hardware-efficient, AI-based classification models for vision tasks, enabling real-time, on-device decision-making to enhance the quality and efficiency of medical examinations.
References
- Paper: “Smart Video Capsule Endoscopy: Raw Image-Based Localization for Enhanced GI Tract Investigation”
- Paper: “Precise Localization Within the GI Tract by Combining Classification of CNNs and Time-Series Analysis of HMMs”
- Paper: “Hardware Accelerator and Neural Network Co-Optimization for Ultra-Low-Power Audio Processing Devices”
Requirements
- Python/PyTorch
- Linux and Git
- Understanding of deep neural networks and basic machine learning concepts
- Successfully attended the lecture “Efficient Machine Learning in Hardware” (recommended)