Hello,
I am having issues reproducing the Axelera Model Zoo benchmarks on a Axelera Metis M.2 module.
My issues seem to be related to pre/post processing, which is why I started with a classification model.
I was hoping to use “fakevideo” as input, to bypass the pre-processing, but that seems to be fixed at 30fps.
Another unresolved issue I have is that opencl is not working on my AMD PC.
For ResNet-50 v1.5, here is what I am getting …
(venv) voyager-sdk$ ./inference.py resnet50-imagenet data/coco --disable-opencl --no-display --show-stats
========================================================================
Element Time(𝜇s) Effective FPS
========================================================================
axinplace-addstreamid0 16 61,066.3
vaapipostproc0 1,927 518.9
videoconvert0 17 56,486.4
axinplace0 7 130,629.7
inference-task0:libtransform_resizeratiocropexcess_0
149 6,677.7
inference-task0:libtransform_totensor_0 7 142,241.8
inference-task0:libinplace_normalize_0 16 59,300.1
inference-task0:libtransform_padding_0 20 47,901.4
inference-task0:inference 2,884 346.7
inference-task0:Inference latency 42,056 n/a
inference-task0:libtransform_paddingdequantize_0
5 184,454.6
inference-task0:libdecode_classification_0 8 116,059.4
inference-task0:Postprocessing latency 726 n/a
inference-task0:Total latency 45,652 n/a
========================================================================
End-to-end average measurement 351.1
========================================================================
Core Temp : 37.0°C
CPU % : 5.5%
End-to-end : 351.1fps
Latency : 42.8ms (min:9.1 max:57.3 σ:3.9 x̄:42.7)ms
