Detectnet for TLT basic model, is there any reason nvidia pick detectnet, not SSD?

I want to handle a real time streaming video, also I want to compare performance of detectnet, SSD and RCNN. I found some resources which compare the performance of SSD, YOLO and RCNN so on, but I couldn’t find the detectnet performance comparing other models. Is detectnet better than any other models? Additionally, does anyone can give me about detectnet’s pros and cons?

For KITTI dataset, detectnet_v2 can get a much better fps than faster-rcnn or SSD.

1 Like

thank you, then is there any dataset can I use rather KITTI? Could you recommand for me?

You can try VOC dataset too. The fps is also the best.