search

enazoe / yolo-tensorrt

TensorRT8.Support Yolov5n,s,m,l,x .darknet -> tensorrt. Yolov4 Yolov3 use raw darknet *.weights and *.cfg fils. If the wrapper is useful to you,please Star it.
MIT License
1.19k stars 316 forks source link

YOLOv5s6 NMS result #123

Open sammilei opened 3 years ago

sammilei commented 3 years ago

Hi,

Thanks for making this repo and easy API. I am trying out the YOLOv5 module with a customized YOLOv5s6 weights. The engines (INT8, FP16, and FP32) can run well, but the NMS result doesn't seem correct. Screenshot from 2021-06-25 16-48-17 Screenshot from 2021-06-25 16-48-36

I tried changing the nms thresh, but it doesn't help. Do you see result like this when testing the diou_nms() function in trt_utils.cpp. Please advise.

Thank you!

positive666 commented 2 years ago

Use tensorrt version 8.2

shining-love commented 2 years ago

Hi,

Thanks for making this repo and easy API. I am trying out the YOLOv5 module with a customized YOLOv5s6 weights. The engines (INT8, FP16, and FP32) can run well, but the NMS result doesn't seem correct. Screenshot from 2021-06-25 16-48-17 Screenshot from 2021-06-25 16-48-36

I tried changing the nms thresh, but it doesn't help. Do you see result like this when testing the diou_nms() function in trt_utils.cpp. Please advise.

Thank you!

hello,have you solved this problem?I have met the same problem.I tried trt7.x、trt8.0 and trt8..2. it don't work!

sunchuanxi commented 2 years ago

i have the same problem,not solve

Nuzhny007 commented 2 years ago

Do you can share data (cfg, weights and image or video)? I'll try to reproduce and debug this case