Please provide the following information when requesting support.
• Network Type Detectnet_v2
• TLT Version: TAO 3.21.11
Previously when using TLT, I used to train most of my Detectnet_v2 detection models with the pretrained detectnet/Resnet 34 because Resnet 18 lacked accuracy and Resnet 50 was too slow.
With a fresh TAO installation !ngc registry model list nvidia/tao/pretrained_detectnet_v2 lacks the Resnet 34 model. Was it removed on purpose? Can I use the “old” resnet 34 pretrained model?