Closed visonpon closed 5 years ago
Hi, We haven't explicitly allowed for this but it should be simple. You should just need to write your own data loader modelled on the VoxCeleb one (so it should return the same order of output) and then load this as opposed to the VoxCeleb in training. Hope that helps.
Hi, thanks for releasing this wonderful work. I'd like to know how to train with my own datasets~ Hope I can receive your response soon~ thanks~