Scene Segmentation using Progressive Learning

Team: SceneSeg

Program: Biomedical Engineering, Computer Science

Scene segmentation is a computer vision task that separates scenes into parts, objects, and concepts. Traditional methods of scene segmentation can use neural networks or random forest models. However, these models tend to perform worse when new data is introduced because of catastrophic forgetting and negative forward transfer. Our goal in this project is to avoid these issues through the use of progressive learning in scene segmentation.