EEGNet is a compact convolutional neural network (CNN) designed for EEG-based brain-computer interfaces (BCIs). The authors introduce the use of depthwise and separable convolutions to construct an EEG-specific model that encapsulates well-known EEG feature extraction concepts. The model is evaluated across four BCI paradigms: P300 visual-evoked potentials, error-related negativity responses (ERN), movement-related cortical potentials (MRCP), and sensory motor rhythms (SMR). The results show that EEGNet generalizes well across paradigms and achieves high performance with limited training data. The model also performs well on both ERP and oscillatory-based BCIs. Additionally, the authors demonstrate three methods to visualize the features learned by EEGNet, enabling interpretation of the learned features. The findings suggest that EEGNet is robust enough to learn a wide variety of interpretable features over a range of BCI tasks, indicating that the observed performance is not due to noise or artifacts in the data.EEGNet is a compact convolutional neural network (CNN) designed for EEG-based brain-computer interfaces (BCIs). The authors introduce the use of depthwise and separable convolutions to construct an EEG-specific model that encapsulates well-known EEG feature extraction concepts. The model is evaluated across four BCI paradigms: P300 visual-evoked potentials, error-related negativity responses (ERN), movement-related cortical potentials (MRCP), and sensory motor rhythms (SMR). The results show that EEGNet generalizes well across paradigms and achieves high performance with limited training data. The model also performs well on both ERP and oscillatory-based BCIs. Additionally, the authors demonstrate three methods to visualize the features learned by EEGNet, enabling interpretation of the learned features. The findings suggest that EEGNet is robust enough to learn a wide variety of interpretable features over a range of BCI tasks, indicating that the observed performance is not due to noise or artifacts in the data.