Best posts made by Viswanadh Chegu

@Jyotish-Kumar i have faced the same issue. the record pipe name has changed in dev branch, try to use
hires_large_encoded or hires_small_encoded based on your resolution requirement. Can find more config parameters in /etc/modalai/voxl-camera-server.conf

Hi @modaltb, @Chad-Sweet

We are facing issue with our own tflite model deployed on voxl2 for text detection bounding boxes.
Here we are seeing "Failed to apply GPU deligate".
From our understanding it seems that model will run on CPU in this case. So we are seeing high inference latency of 7-8 seconds, which we want to minimize by running it on GPU.
Can you please help use with this issue.

PS: Our tflite model size is 41MB.

Attaching snapshot from voxl2 run for reference.

Thanks