Config file path:/home/sckj/yolov8/DeepStream-Yolo-master/config_infer_primary_ yolov8.txt,NvDsInfer Error:NVDSINFER CONFIG FAILED

[1354065425]

ferContextImpl::generateBackendContext() [UID = 1]: build backend context failed 0:17:27.919840104 15844 0xaaaaefab6300 ERROR nvinfer gstnvinfer.c pp:674:gst_nvinfer_logger: NvDsInferContext[UID 1]: Error in NvDsIn ferContextImpl::initialize() [UID = 1]: genera te backend failed, check config file settings 0:17:27.919939849 15844 0xaaaaefab6300 WARN nvinfer gstnvinfer.c pp:888:gst_nvinfer_start: error: Failed to create NvDsInferContext instance 0:17:27.919998569 15844 0xaaaaefab6300 WARN nvinfer gstnvinfer.c pp:888:gst_nvinfer_start: error: Config file path: /home/sckj/yolov 8/Deepstream-Yolo-master/config_infer_primary_yoloV8.txt，NvDsInfer Error: NVDSI NFER CONFIG FAILED ** ERROR: : Failed to set pipeline to PAUSED Quitting

[1354065425]

[Today-fine]

Same error with my model using V5, working on it

[1354065425]

my model is yolov8，follow the guidance，but not work

[Today-fine]

I have no problem switching to Orin according to the official document, but I can’t switch to Xavier

[Today-fine]

my model is yolov8，follow the guidance，but not work

I found the problem, unplugged the monitor, and then changed it to file output to generate the engine file, but my device has a problem and the power is cut off when playing the video, OP can try

[marcoslucianops]

0:17:27.919998569 15844 0xaaaaefab6300 WARN nvinfer gstnvinfer.cpp:888:gst_nvinfer_start: error: Config file path: /home/sckj/yolov8/Deepstream-Yolo-master/config_infer_primary_yoloV8.txt，NvDsInfer Error: NVDSINFER CONFIG FAILED

Try to use the model exported without the --dynamic flag.

[1354065425]

0:17:27.919998569 15844 0xaaaaefab6300 WARN nvinfer gstnvinfer.cpp:888:gst_nvinfer_start: error: Config file path: /home/sckj/yolov8/Deepstream-Yolo-master/config_infer_primary_yoloV8.txt，NvDsInfer Error: NVDSINFER CONFIG FAILED

Try to use the model exported without the --dynamic flag.

ok, the model exported without the --dynamic flag, but not detection

[1354065425]

python3 gen_wts_yoloV8.py -w yolov8s.pt -s 640 cp labels.txt Deepstream-yolo cp yolov8s.onnx Deepstream-yolo cd Deepstream-yolo && CUDA_VER=11.4 make -C nvdsinfer_custom_impl_Yolo deepstream-app -c deepstream_app_config.txt for model_b1_gpu0_fp32.engine, but not detect this videos

[marcoslucianops]

The wts method is outdated and isn't supported anymore. Please check the new docs for export method.

[1354065425]

The wts method is outdated and isn't supported anymore. Please check the new docs for export method.

i use the command python3 export_yoloV8.py -w yolov8s.pt -s 640 not wts method

[marcoslucianops]

The last comment was python3 gen_wts_yoloV8.py -w yolov8s.pt -s 640. That's why I said that.

[marcoslucianops]

Please send here the output from the terminal when you run the deepstream-app.

[1354065425]

Please send here the output from the terminal when you run the deepstream-app.

[marcoslucianops]

Please send the full output. I would like to see the output layers/shapes.

[1354065425]

Please send the full output. I would like to see the output layers/shapes

[marcoslucianops]

It seems ok. Can you send the deepstream_app_config.txt and the config_infer_primary_yoloV8.txt files?

[1354065425]

It seems ok. Can you send the deepstream_app_config.txt and the config_infer_primary_yoloV8.txt files?

Okay, thank you for your reply This is my deepstream_app_config.txt and config_infer_primary_yoloV8.txt config_infer_primary_yoloV8.txt deepstream_app_config.txt

[marcoslucianops]

Please change

#net-scale-factor=0.0039215697906911373

to

net-scale-factor=0.0039215697906911373

in the config_infer_primary_yoloV8.txt file.

[1354065425]

Please change

#net-scale-factor=0.0039215697906911373

to

net-scale-factor=0.0039215697906911373

in the config_infer_primary_yoloV8.txt file.

no decetion

[marcoslucianops]

Can you test the official YOLOv8s or other variant with uncomented net-scale-factor=0.0039215697906911373? Did you add normalization to your model in the training?

[1354065425]

Can you test the official YOLOv8s or other variant with uncomented net-scale-factor=0.0039215697906911373? Did you add normalization to your model in the training?

Okay, I'll try yolov8s now

[1354065425]

Can you test the official YOLOv8s or other variant with uncomented net-scale-factor=0.0039215697906911373? Did you add normalization to your model in the training?

The steps are all based on your document, but even if it is not recognized, the video can still be played out

[marcoslucianops]

The YOLOv8 needs the net-scale-factor=0.0039215697906911373. You need to change it only if your model is trained with image normalization.

[1354065425]

The YOLOv8 needs the net-scale-factor=0.0039215697906911373. You need to change it only if your model is trained with image normalization. yolov8s Still the Same

[marcoslucianops]

Did you delete the old engine file to run the YOLOv8s?

[1354065425]

Did you delete the old engine file to run the YOLOv8s?

yes，I removed previous engine

[1354065425]

Did you delete the old engine file to run the YOLOv8s?

I still haven't found where the problem lies

[marcoslucianops]

Hi, send the ONNX model to my email please.

[1354065425]

Hi, send the ONNX model to my email please.

Okay, I sent you an email,My email name is wangjin0992@gmail.com

[1354065425]

Hi, send the ONNX model to my email please.

Hey, brother, I tried using YOLOV7 and it's okay. It can recognize it normally, but YOLOV8 doesn't recognize it. Is the network structure of YOLOV8 different? Is the head and network separate?

[marcoslucianops]

What's your PyTorch version?

[1354065425]

What's your PyTorch version?

torch version is 1.14

[marcoslucianops]

It should work with PyTorch 1.14. The network is different, but the code is ready to work with the YOLOv7 and YOLOv8 when exported with the export file available on this repo.