I have switched to tensorrt from cpu detection about 2 days ago. I am not getting any object detection now. I ahve recently needed to fix other problems as well including cache size and shm size, as well as birdseye settings. I luckily have frigate connected to home assistant which provides a fair amount of historical information. It seems like motion is working as seen here:
But object detection seems to be stuck in time since when I switched to tensorrt as seen here:
I have tried modifying different thresholds with no changes, at the moment everything is almost back to default settings as you will see in the config. I am not sure what I did wrong here. As suggested I waited a while to let log files form, however after about an hour and a half they are still fairly slim. Something I couldn't have said a few days ago but I bet makes troubleshooting hard now.
2023-02-23 18:10:19.263341007 [2023-02-23 18:10:19] frigate.app INFO : Starting Frigate (0.12.0-27a31e7)
2023-02-23 18:10:19.292887845 [2023-02-23 18:10:19] peewee_migrate INFO : Starting migrations
2023-02-23 18:10:19.326871339 [2023-02-23 18:10:19] peewee_migrate INFO : There is nothing to migrate
2023-02-23 18:10:19.341440899 [2023-02-23 18:10:19] detector.tensorrt INFO : Starting detection process: 980
2023-02-23 18:10:19.352220207 [2023-02-23 18:10:19] frigate.app INFO : Output process started: 982
2023-02-23 18:10:19.352732530 [2023-02-23 18:10:19] frigate.app INFO : Camera processor started for Front_Door: 986
2023-02-23 18:10:19.352792883 [2023-02-23 18:10:19] frigate.app INFO : Camera processor started for Front_Yard: 987
2023-02-23 18:10:19.357996800 [2023-02-23 18:10:19] frigate.app INFO : Camera processor started for Shed_Camera: 991
2023-02-23 18:10:19.362446943 [2023-02-23 18:10:19] frigate.app INFO : Camera processor started for Driveway_Doorbell: 992
2023-02-23 18:10:19.367259005 [2023-02-23 18:10:19] frigate.app INFO : Camera processor started for Living_Room: 994
2023-02-23 18:10:19.370818458 [2023-02-23 18:10:19] frigate.app INFO : Camera processor started for Barn: 995
2023-02-23 18:10:19.376387884 [2023-02-23 18:10:19] frigate.app INFO : Camera processor started for Septic_Shed: 997
2023-02-23 18:10:19.380107259 [2023-02-23 18:10:19] frigate.app INFO : Camera processor started for Septic_Driveway: 998
2023-02-23 18:10:19.384066445 [2023-02-23 18:10:19] frigate.app INFO : Camera processor started for Septic_Inside: 1000
2023-02-23 18:10:19.387580523 [2023-02-23 18:10:19] frigate.app INFO : Capture process started for Front_Door: 1003
2023-02-23 18:10:19.391938361 [2023-02-23 18:10:19] frigate.app INFO : Capture process started for Front_Yard: 1006
2023-02-23 18:10:19.396560138 [2023-02-23 18:10:19] frigate.app INFO : Capture process started for Shed_Camera: 1011
2023-02-23 18:10:19.403029742 [2023-02-23 18:10:19] frigate.app INFO : Capture process started for Driveway_Doorbell: 1017
2023-02-23 18:10:19.408386984 [2023-02-23 18:10:19] frigate.app INFO : Capture process started for Living_Room: 1023
2023-02-23 18:10:19.413486304 [2023-02-23 18:10:19] frigate.app INFO : Capture process started for Barn: 1028
2023-02-23 18:10:19.419914754 [2023-02-23 18:10:19] frigate.app INFO : Capture process started for Septic_Shed: 1034
2023-02-23 18:10:19.426373686 [2023-02-23 18:10:19] frigate.app INFO : Capture process started for Septic_Driveway: 1040
2023-02-23 18:10:19.432466826 [2023-02-23 18:10:19] frigate.app INFO : Capture process started for Septic_Inside: 1044
2023-02-23 18:10:19.642305979 [2023-02-23 18:10:19] frigate.detectors.plugins.tensorrt INFO : [MemUsageChange] Init CUDA: CPU +321, GPU +0, now: CPU 408, GPU 284 (MiB)
2023-02-23 18:10:19.647814040 [2023-02-23 18:10:19] frigate.detectors.plugins.tensorrt INFO : Loaded engine size: 12 MiB
2023-02-23 18:10:20.162049560 [2023-02-23 18:10:20] frigate.detectors.plugins.tensorrt INFO : [MemUsageChange] Init cuBLAS/cuBLASLt: CPU +855, GPU +358, now: CPU 1298, GPU 658 (MiB)
2023-02-23 18:10:20.244360102 [2023-02-23 18:10:20] frigate.detectors.plugins.tensorrt INFO : [MemUsageChange] Init cuDNN: CPU +126, GPU +58, now: CPU 1424, GPU 716 (MiB)
2023-02-23 18:10:20.245274829 [2023-02-23 18:10:20] frigate.detectors.plugins.tensorrt INFO : [MemUsageChange] TensorRT-managed allocation in engine deserialization: CPU +0, GPU +11, now: CPU 0, GPU 11 (MiB)
2023-02-23 18:10:20.247670400 [2023-02-23 18:10:20] frigate.detectors.plugins.tensorrt INFO : [MemUsageChange] Init cuBLAS/cuBLASLt: CPU +0, GPU +8, now: CPU 1411, GPU 708 (MiB)
2023-02-23 18:10:20.247719137 [2023-02-23 18:10:20] frigate.detectors.plugins.tensorrt INFO : [MemUsageChange] Init cuDNN: CPU +0, GPU +8, now: CPU 1411, GPU 716 (MiB)
2023-02-23 18:10:20.247740309 [2023-02-23 18:10:20] frigate.detectors.plugins.tensorrt INFO : [MemUsageChange] TensorRT-managed allocation in IExecutionContext creation: CPU +0, GPU +3, now: CPU 0, GPU 14 (MiB)
2023-02-23 18:11:27.927567131 [2023-02-23 18:11:27] frigate.record WARNING : Unable to find file from recordings database: /media/frigate/recordings/2023-02-03/06/Driveway_Doorbell/05.41.mp4
You are using the 288 model, but you have set the model size to 416 so this is incompatible. The model size needs to be correct for the model that is being used
Describe the problem you are having
I have switched to tensorrt from cpu detection about 2 days ago. I am not getting any object detection now. I ahve recently needed to fix other problems as well including cache size and shm size, as well as birdseye settings. I luckily have frigate connected to home assistant which provides a fair amount of historical information. It seems like motion is working as seen here:
But object detection seems to be stuck in time since when I switched to tensorrt as seen here:
I have tried modifying different thresholds with no changes, at the moment everything is almost back to default settings as you will see in the config. I am not sure what I did wrong here. As suggested I waited a while to let log files form, however after about an hour and a half they are still fairly slim. Something I couldn't have said a few days ago but I bet makes troubleshooting hard now.
Version
0.12.0-27A31E7
Frigate config file
Relevant log output
FFprobe output from your camera
Frigate stats
Operating system
UNRAID
Install method
Docker Compose
Coral version
Other
Network connection
Wired
Camera make and model
Amcrest, Reolink
Any other information that may be helpful
NVIDIA SMI Snapshot