I am planning on buying the Jetson AGX Xavier. The benchmarks in the official product website are for image segmentation and image classification. Is it possible to get the inference benchmarks for different object detection models viz Detectnetv2-resnet10,Detectnetv2-resnet18,Tiny-Yolo v3 and SSD-Mobilenetv2. Can you also list the FPS for each model?
Hi sivaishere96, you can find some additional benchmarks for Jetson AGX Xavier through the links below:
Thanks dusty_nv. In the mlperf results, can you explain the results of multi-stream. Does it mean the AGX Xavier processes 102 images per 50 milliseconds for SSD MobileNet-v1?
That’s correct (you can see the multi-stream latency constraints for the different benchmarks here).
One more clarification dusty_nv. From the multi-stream mlperf results can we infer that we can run SSD-Mobilenetv2 on 102 video steams at 20fps (1000ms/50ms) using deepstream?
can you tell me what is the resolution of the frame for yolov3 in that benchmark