yolov5_gpu_optimization icon indicating copy to clipboard operation
yolov5_gpu_optimization copied to clipboard

Published 20 hours ago • verified publisher icon NVIDIA-AI-IOT

mAP decreased when run on Jetson using INT8 model

Open nghoaithuong opened this issue 2 years ago • 3 comments

When I tried to run INT8 model on Jetson, mAP drop ~10%. Model INT8 can't detect small object when FP16 model was normal. I used newest docker TensorRT nvcr.io/nvidia/l4t-tensorrt:r8.4.1.5-devel

nghoaithuong avatar Dec 19 '22 04:12 nghoaithuong

Did you run with TensorRT sample ? What's the model you used ? And did you try int8 on dGPU ?

Tyler-D avatar Dec 19 '22 06:12 Tyler-D

Yes, I run TRT sample with my custom model using int8 on GPU. This issue only occurred when I ran on jetson, it's normal when ran on PC with above docker.

nghoaithuong avatar Dec 19 '22 07:12 nghoaithuong

OK. Then it is potentially a L4T TensorRT issue. So my suggestion is that you can try to pick some layers and set the precision to be fp16.

Tyler-D avatar Dec 20 '22 03:12 Tyler-D