The outspread of COVID-19 significantly effected both the world economy and public health. Motivated by the collaborative efforts of the free resource community in compiling COVID-19 data and the successful utilization of deep learning techniques in disease identification, the aim is to create a novel COVID-19 detection system that used images from chest X-rays. We designed an advanced Convolutional Neural Network (CNN) architecture, building upon the EfficientNet-B0 model, to detect COVID-19 with high accuracy. This enhanced model incorporates a compound coefficient to precisely balance depth, breadth, and resolution. By employing predefined scaling coefficients, the network’s dimensions are uniformly adjusted. However, fine-tuning the hyperparameters of EfficientNet-B0 is essential for optimizing its performance. Thus, we propose an automated hyperparameter optimization using the Salp Swarm Algorithm (SSA) to fine-tune EfficientNet-B0 for COVID-19 classification. Our SSA-EfficientNet-B0 model outperforms Particle Swarm Optimization and Genetic Algorithm. The optimized neural network achieves a remarkable accuracy rate of 98.89% in detecting COVID-19 positive cases, surpassing alternative methods.