What is dropout and how is it used in computer vision and deep learning networks? How does dropout help prevent overfitting during neural network training? What happens to neurons during the dropout process? How does dropout improve model generalization and performance on unseen data? What are the advantages and limitations of using dropout in computer vision models?