Accurate identification of visually similar fish species in post-harvest environments remains a challenging task due to occlusions, inter-species similarity, and complex backgrounds. This study introduces a real-world dataset comprising 552 RGB images of red snapper and colorado snapper, captured under diverse illumination and operational conditions across multiple fishery areas. We propose a lightweight object detection framework based on the YOLOv8n architecture, designed for robust detection and discrimination of the two species. Experimental results demonstrate that YOLOv8n achieves a mean Average Precision (mAP@0.5) of 81.5% and outperforms larger YOLOv8 variants in convergence stability and recall, while maintaining computational efficiency suitable for real-time deployment. This work highlights the potential of deep learning-based detectors for operational fishery monitoring and the identification of closely related species under challenging real-world conditions.