The manufacturing of semiconductors is becoming increasingly expensive and complex, while market demand continues to grow. This makes the automation and optimization of the production process more crucial. The inspection stage is a critical point that is susceptible to improvement, as it still relies on visual inspection by specialized technicians. This results in increased production costs and time, as well as a higher risk of errors due to fatigue and bias affecting the performance of inspection personnel. In this context, the design and implementation of a new automatic inspection algorithm for detecting missing solder balls in encapsulated semiconductors is presented. This algorithm utilizes computer vision and deep learning. The images used for the training and testing of the models come from semi-automated inspection equipment that is part of the production process, and the obtained results demonstrate the feasibility of integrating such algorithms in a realistic production environment.