markusbr

Do you use Intel-CPU and Intel-GPU in your environment...? Can you check with `lscpu`? Maybe your SoC is too older or too new.

Can you check with `lspci` as well, like `lspci | grep -i vga` whether your SoC has an Intel GPU at all (I haven`t google'd your CPU i7-9700)?

Hmmm, then you can't use Intel-based (vaapi) gstreamer plugins like `vaapih264dec`... Do you know NVidia-related video-codec-gstreamer-plugins? Or you use pure SW/pure CPU-video-codecs and gstreamer plugins, like libav-based.

Maybe the iGPU is disabled in BIOS? Or you use a SOC-variant (early sample, engineering-sample) where the iGPU got disabled? If it doesn't appear on the PCI-bus, then it's not...

Something like this: ``` export DISPLAY=:0 mkdir /tmp/cl_cache export cl_cache_dir=/tmp/cl_cache cd myworkingdir gst-launch-1.0 filesrc location=myvideo.h264 ! h264parse ! vaapih264dec ! gvadetect model=mymodel.xml device=GPU model-proc=mymodelproc.json ! gvametaconvert ! gvawatermark ! xvimagesink...

Can you provide more details about how you have set-up your environment, by following the installation-description under "https://dlstreamer.github.io/get_started/install/install_guide_ubuntu.html"? You might need to get dependencies (like Kafka or MQTT) installed first...

@howarwang which YoloV5 model have you tested? Have you tested `yolov5s`? Can you provide details and share what you did and how it looked like, please?

@howarwang yes, the model is working with the YoloV5 Github repo's test-file `detect.py`. Could you check whether the model can be used with DL-Streamer, too, please?

You might find additional information in another issue/question like this: "https://github.com/dlstreamer/dlstreamer/issues/178" ("How to attach and extract custom metadata generated by custom written plugin?")

Are you sure `vainfo` on the HOST shows the dGPU and not the iGPU (if your used SoC has an integrated, embedded GPU)? Does `lspci | grep VGA` shows only...