question regarding the provided pretrain models

[isalirezag]

Are the provided pretrained models fine-tune on VOC dataset or they are just pretrained on Imagenet?
e.g. I am talking about "ImageNet-pretrained networks: "