How can i get 500 FPS on object detection using t4, dgpu and DS version5.1. I am not getting so

DS 5.1
tensor rt 7

Can you share more on your use case? Which DS sample do you run?

Ya sure, i want to use any object detection model with 10 cameras with 500 fps minimum in deepstream test test 3 app. it will be great if you give me code running with 500 fps in test3 with 10 cameras(2000*2000 resolution). you can share me weights and other file on google drive or somewhere else. so that i only need to run it. Thanks in advanced.

Do you meet any issue to run DS test 3 app?

Ya i have run test3 without any issue but i want use yolo v4 or may be yolo v5 to achieve high accuracy(it is also very important) for custom object detection.

with GitHub - NVIDIA-AI-IOT/yolov4_deepstream (you can find the sample with onnx under google drive here) .

The DeepStream pipeline perf on T4 is. If you run it with INT8, you may could get ~450 fps perf. But to run it with INT8, you need generate the INT8 calibrtion table based on your own image data - Developer Guide :: NVIDIA Deep Learning TensorRT Documentation


This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.