DS4.0.2 TX2 : precision detection decreased a lot after update yoloV3 network model

[guibli]

Hello,
I updated yolov3 model based on DS4.0.2 ,and modified yolo parse code, then tensorRT could parse the net successfully ,but the precision detection decreased a lot, I double check the yolo parse code and all the config file ,not found any problem.

I updated yolov3 model network according yoloV3-spp and Slim yoloV3 ,both decreased a lot of precision detection.

I have created this topic last year ,but didn’t solved until now, for detail please see

DS4.0 TX2 : precision detection decreased a lot

Hello, I’m using slim yolov3 model based on DS4.0 ,and modified yolo parse code, tensorRT could parse the net successfully ,but the precision detection decreased a lot, I try to use diffferent network-mode ,but it still no improvement. I checked all the config file ,not found any problem. I just update two points based on source code of createYoloNetwork function. Could you help me to check this problem ,I think it shouldn’t decrease so much for precision. adding parse more tensor for “…

Environment

Jetpack 4.2.2
TX2
DS4.0.2
project based on objectDetector_Yolo

It’s very import for me ,I can’t sure where is the issue points.
Thanks for your support .

[SunilJB]

Moving to Jetson forum so that Jetson team can take a look.

[guibli]

thanks a lot

[mchi]

Hi @guibli,
Do you mean the accuracy drop is caused by deepstream?

BTW, we just released a project that demonstrate how to run TLT2.0 models on DeepStream - GitHub - NVIDIA-AI-IOT/deepstream_tao_apps: Sample apps to demonstrate how to deploy models trained with TAO on DeepStream, would you take a look if the YoloV3 in this project can help you?

[guibli]

OK,I will try ,thanks