Pre-trianed Backbones trained on which dataset?

The backbones provided within TLT (resnet, mobilenet) - which datasets are they pretrained on?

Do you mean the pretrained models?
See https://ngc.nvidia.com/catalog/models/nvidia:tlt_pretrained_object_detection and
https://ngc.nvidia.com/catalog/models/nvidia:tlt_pretrained_detectnet_v2

It is trained on a subset of the Google OpenImages dataset. Following backbones are supported with these detection networks.

resnet10/resnet18/resnet34/resnet50/resnet101
vgg16/vgg19
googlenet
mobilenet_v1/mobilenet_v2
squeezenet
darknet19/darknet53

thanks