Hi,i’m so sorry to bother you, i have some question to you:
1、I see in the NGC model that the size of the pretrained model for object detection and image classification is the same, are these files the same? and What tasks were used to train these these models（i，mobilenet_v2，resnet50，cspdarknet53） on the Google OpenImages dataset? Classification or object detection?
Uploading: image.png…
2. What is the effect of resolution on the final model accuracy in TAO training? Does the resolution setting have to be square? and rectangle is ok?
3、The tao trained model is deployed in deepstream, how to set the resolution to get the best detection effect？
I look forward to your reply thank you