Tiered Object Detection in Underwater Environments

This research proposes using a tiered detection pipeline to improve computer vision models that must perform across a variety of environments. The tiered pipeline classifies the environment, then uses the corresponding detection model. This method will be explored in the context of underwater navigation with different bodies of water as the varying environment. Data from the two environments is used to train models for both classification and detection, which is then compared against a single model trained on data from both environments. Comparing these two pipelines, no significant differences are apparent.