We need libBlas and libLAPACK for deep learning via tensorflow, and are not using mlkit because of its feature limitations. We are noticing an intermittent 8 second lag on models that usually take around half a second to complete.
During that time it usually reports that "[Technique] World tracking performance is being affected by resource constraints"
Now this could be tensorflow itself, or something I am not thinking about inside our code, but I am just asking whether these two running at the same time might cause an issue or if anyone else has noticed it.
If so, how can it be remedied?
We've seen this as well. But on ARKit 1.5 (Beta5/Beta6). Are you on the beta?
We moved off of it and it no longer halts for long periods of time. However, our model normally takes 0.4 seconds without ARKit running and takes 1.7 to 2.5 when it is running.