I reckon that increasing the neural network output’s resolution by adding extra deconvolutional layers was a common practice, but I couldn’t find any paper or blog post talking about that. Does anyone know about related bibliography?
Thank you in advance.