Renewable energy significantly contributes to power generation to meet the growing energy demand. It is sourced from solar, wind, hydroelectric, and other sources. Among these, solar energy is considered the most suitable due to its cleanliness and its ability to directly convert sunlight into electrical power through solar photovoltaic (PV) modules. One of the biggest challenges with solar panels is the random fluctuation of their power output due to variations in irradiance. The concept of maximum power point tracking (MPPT) techniques has been introduced to address this non-linear behavior of PV systems and optimize their efficiency. Various MPPT techniques have been proposed, based on both conventional and intelligent methods. In this work, a novel image processing-based MPPT technique is introduced to enhance the efficiency of PV systems. The irradiance level is accurately classified using the self-learned EfficientNetB0 deep learning model. The parameters of the EfficientNetB0 model are adjusted using Tuna Swarm Optimization. The results show that the tracking efficiency is higher compared to other intelligent MPPT techniques. The classification accuracy of the proposed learning model surpasses that of conventional models.