This study puts forward a dual-frequency hybrid pooling framework that integrates the Fast Fourier Transform (FFT) and the Discrete Wavelet Transform (DWT) within a modified VGG-16 architecture to improve feature preservation and reduce computational complexity. The framework was evaluated using the PlantVillage tomato leaf dataset, which contains 25,851 images across 11 classes. Experimental results showed that the proposed model attained 99.77% classification correctness with substantially fewer convolutional layers than the conventional VGG-16. The results show that duo-frequency hybrid pooling improves feature representation, enhances computational efficiency, and provides a scalable deep learning CNN framework for exact agriculture applications.