This paper presents an automated apple classification system using near-infrared (NIR) imaging and novel lightweight convolutional neural networks (CNNs) to distinguish between apples in rotten, good, and bruised condition. The methodology relies on processing NIR images to capture subtle differences in the texture and composition of apple surfaces, which enables precise classification through the proposed network architecture. The algorithm optimizes identifying characteristic patterns associated with each category, allowing for improved accuracy and speed. Comparative results show that this approach outperforms state-of-the-art methods in efficiency, with potential applications in automated sorting systems and quality control processes.