Zebra and quagga mussels, invasive species in North America, cause significant ecological and economic damage by out-competing native species and obstructing water infrastructure. Traditional methods for detecting mussel larvae are costly and time-consuming, necessitating the development of automated monitoring procedures and super-resolution (SISR) techniques to enhance low-resolution water sample images. This study investigates the effectiveness of three advanced SISR methods—Super-Resolution Convolutional Neural Network (SRCNN), Enhanced Super-Resolution Generative Adversarial Networks (ESRGAN), and Water-Net—across three classification frameworks: a baseline model, a supervised contrastive learning model, and a convolutional neural network (CNN) in improving the classification accuracy of invasive and non-invasive larvae. By integrating these SISR techniques with supervised contrastive learning, the aim was to enhance image quality and feature representation. The results demonstrate that ESRGAN significantly outperforms SRCNN and Water-Net, achieving the highest classification accuracy of 96.96% due to its advanced architecture and feature enhancement capabilities. These findings underscore the critical role of image quality in species classification and highlight the potential of ESRGAN in supporting effective ecological management and mitigation strategies for invasive species.