Extracting meaningful information from Electroencephalogram (EEG) signals in the human brain using deep learning techniques is an increasingly popular research topic. In particular, identifying visual stimuli from EEG signals has significant potential in brain-computer interface (BCI) applications. In this work, we use an Oscillatory Convolution Neural Network (OCNN) to decode visual stimuli from EEG data. The system consists of two main components: a classifier and a decoder. The classifier is implemented using OCNN, which incorporates hidden layers composed of nonlinear neural Hopf oscillators. The features extracted from the final layer of the classifier are then fed into a decoder, a conditional Generative Adversarial Network (cGAN). The model is trained on the ThoughtViz EEG dataset and demonstrates superior classification performance compared to existing methods reported in the literature. We achieved the highest classification accuracy of 86.19% on the character EEG dataset and 89.03% on the image EEG dataset, outperforming existing models. Furthermore, on the decoder side, the reconstructed images also yielded superior classification accuracy and SSIM score compared to previously proposed methods in the literature, demonstrating the effectiveness of our feature extraction and reconstruction pipeline.