PP-YOLOE on DeepStream

[marcoslucianops]

Repository to use PP-YOLOE model on NVIDIA DeepStream SDK

Link: https://github.com/marcoslucianops/DeepStream-Yolo

Config

board = NVIDIA Tesla V100 (AWS: p3.2xlarge)
network-mode = FP16
batch-size = 1
sample = 1920x1080 video

NMS config

iou-threshold=0.3
score-threshold=0.25
topk=300

PP-YOLOE-s (ppyoloe_crn_s_400e_coco) performance comparison

	DeepStream	Paddle TensorRT
FPS	393.71	333.3