Scleractinian corals are essential reef-building organisms that support marine biodiversity, fisheries productivity, and coastal protection. However, they are increasingly threatened by climate change, coral bleaching, ocean acidification, pollution, and destructive human activities—leading to widespread degradation of reef ecosystems. Accurate detection and monitoring of these corals are critical for understanding reef health and informing conservation strategies. A widely adopted technique for coral assessment is the quadrat method, which involves capturing standardized photographic data within fixed-area frames to document coral cover and distribution. This study explores the use of a deep learning-based object detection approach using the Real-Time Detection Transformer (RT-DETR) to identify Scleractinian corals in underwater quadrat images. The dataset consisted of expertly annotated images representing various coral morphologies. Image preprocessing and augmentation processes were carried out to enhance model robustness under challenging underwater conditions. The RT-DETR model was trained over 200 epochs with early stopping and evaluated using standard detection metrics. The model achieved strong performance, with a best mAP@0.50 of 0.8421, mAP@0.75 of 0.6680, and mAP@0.50:0.95 of 0.5781. Per-class evaluation showed high detection accuracy for Encrusting, Massive, and Non-acropora Branching corals, while lower recall was noted for visually subtle categories such as Acropora Branching and Submassive. Overall model agreement was high, with a Matthews Correlation Coefficient (MCC) of 0.6570, Cohen’s Kappa of 0.6513, and Balanced Accuracy of 0.6637. These findings emphasize the potential of RT-DETR to support conservation-driven coral monitoring efforts, providing a scalable and efficient tool for protecting vulnerable reef ecosystems.