Description of Invention
This study leverages deep learning for coral species identification, focusing on convolutional neural networks like EfficientNetV2, MobileNetV3, and ConvNeXt. Utilizing a diverse dataset from the Moorea Coral Reef LTER site and CoralNet, EfficientNetV2 emerged as the most accurate and efficient model. The research then refined EfficientNetV2 using techniques like early training halting and image proportion consideration. The enhanced model showed exceptional precision, validating its utility for coral conservation. This approach offers a resource-efficient tool to advance coral conservation efforts.