Detecting, monitoring, and conserving marine species within ecosystems is essential to preserve environmental balance. Traditional methods rely on manual detection of species on the ocean floor, a process that is both labor-intensive and time-consuming. These methods often require extensive sampling efforts, which can be intrusive and potentially harmful to marine environments. Moreover, these limitations hinder large-scale, continuous monitoring, reducing the capacity to effectively address changes in marine ecosystems. To overcome these challenges, we present a new MSD-YOLOViT10 model, which integrates a visual transformer with the YOLOv10 neck. This integration enhances the representation of marine species, leading to significant improvements in detection performance within marine ecosystems. Preliminary results on the IITGoa Marine Vision (IITGoaMV) dataset demonstrate that MSD-YOLOViT10 outperforms existing state-of-the-art methods in marine species detection and recognition. The resources will be available at https://github.com/mkapoor1910/Marine-Species-Detection .