Confusion matrix representing the accuracy, precision, recall, specificity, negative predicted values, and F1-score for audio, video, and ensemble neural network models on a testing set of 80 videos.

We also report on two other models for which we take the final prediction as ASD if either audio or video model predicts a subject as ASD (OR Model) and we take the final prediction as ASD only if both models predict a subject as ASD (AND model).