This repository supports the paper titled: "Kelp Wanted: Segmenting Kelp Forests", which was written following participation in the Semantic Segmentation Challenge organized by Driven Data.
The full paper detailing our approach, methodology, and results is available in this repository. Please refer to the paper for an in-depth understanding of our work:
Paper: Kelp Wanted: Segmenting Kelp Forests
One of my main contributions besides developing pipeline and analyzing dataset of this work is the introduction of anchor boxes to highlight specific areas of importance within the images. This technique was instrumental in improving the segmentation accuracy by focusing on relevant regions of interest, thereby reducing noise and enhancing feature extraction.
We acknowledge Driven Data for organizing the competition and providing the dataset used in this study.