Enable the vvas_xinfer profiler when the pipeline contains inference:
gst-launch-1.0 \
<pipeline-before-xinfer> ! \
vvas_xinfer config-file=<config.json> \
enable-profiler=true \
profiler-file=/tmp/xinfer_profile.json \
profiler-log-interval=5 ! \
<pipeline-after-xinfer>
The profiler JSON reports the selected backend, average FPS, and timing for
preprocess, inference, and postprocess stages. Use this output to
check whether ML inference is the bottleneck or whether time is being spent in
pre-processing or post-processing. For the full profiler description, see
Profiler Support.