CodeProject AI is Crashing Windows 10 with Blue Iris 5
Posted: Mon Dec 18, 2023 4:02 pm
Hi all. I have had weeks of pain with Blue Iris, lockups and blue screens/hard freezes due to CodeProject AI.
Symptoms
My computer randomly locks up/reboots and there doesn’t seem to be any reason why other than Blue Iris/CodeProject Server windows service. If I turn off Blue Iris via mmc/services console (stop the service) I do not get random crashes and the system is stable. I assumed this must have something to do with the decoding of the camera streams. I debugged for days by finally turning off the camera streams and CodeProject AI, I turned everything back on one by one, ensuring I'm only using H.264 streams, and so far, WITHOUT CodeProject AI running, it has been stable and not crashing for a couple of days which is a record for me.
I did have CodeProject setup to detect persons/cars in certain areas. The bummer for me is that it was working PERFECTLY. I absolutely love the integration. I wasn't getting tons of false alerts from my cat/dog, I was only getting notified via email when I wanted to be. Now I'm wondering if anyone has had this issue, how I might go about debugging CodeProject to solve it? With these hard freezes/random lockups I don't have any good stacktrace to go from.
System.
Blue Iris 5.8.1.5
i7 8700K Processor @ stock speed with integrated graphics
Nvidia GTX 1060 3GB (Asus)
Motherboard: Gigabyte Z370 AORUS Gaming 5
32GB DDR4 at 2400mhz stock speed
Samsung 980 Evo SSD 512GB (plenty of free space)
Multiple other internal drives for various reasons
Windows 10 latest updates/edition
Nvidia latest graphics driver direct from Nvidia.com (clean)
Installed most recent network card (Intel ethernet) drivers (clean)
Installed most recent intel graphics and chipset drivers (clean)
Cameras (4x with PoE switch)
Amcrest 5MP Turret POE Camera, UltraHD Outdoor IP Camera POE with Mic/Audio, 5-Megapixel Security Surveillance Cameras, 98ft NightVision, 103° FOV, MicroSD (256GB), (IP5M-T1179EW-28MM)
Camera settings (double checked on every camera and tried multiple different settings)
Main Stream
H.264
Smart Code: Off
1920x1080 resolution (even though the cameras can do 4k)
25FPS
VBR
Quality: 4
Max Bit Rate: 4096
Sub Stream
H.264
704*480(D1) resolution
15FPS
CBR
Bit Rate: 512Kb/S
Frame Interval: 30
Within Blue Iris, I have tried setting up the cameras to decode with “Intel”, “Intel + VPP”, “Nvidia”. and CPU only with NO hardware acceleration, the system will lock up and crash regardless of which is chosen if I have AI enabled. With Intel + VPP, sometimes I might get lucky and it might go nearly a full day without crashing, but it still crashes every day.
It’s obvious that something is not playing nicely with Blue Iris/CodeProject integration, I thought maybe it was the camera streams/encoding, but they are all H.264 and fairly easy bitrate at 1080p…it should not have any issue with 4 cameras at 1080p.
Any suggestions? If not, do you have any other thoughts or settings you might suggest I change to try to achieve normal system stability? I would absolutely love to use the AI integration as it makes my alerts actually useful.
Thank you!
Symptoms
My computer randomly locks up/reboots and there doesn’t seem to be any reason why other than Blue Iris/CodeProject Server windows service. If I turn off Blue Iris via mmc/services console (stop the service) I do not get random crashes and the system is stable. I assumed this must have something to do with the decoding of the camera streams. I debugged for days by finally turning off the camera streams and CodeProject AI, I turned everything back on one by one, ensuring I'm only using H.264 streams, and so far, WITHOUT CodeProject AI running, it has been stable and not crashing for a couple of days which is a record for me.
I did have CodeProject setup to detect persons/cars in certain areas. The bummer for me is that it was working PERFECTLY. I absolutely love the integration. I wasn't getting tons of false alerts from my cat/dog, I was only getting notified via email when I wanted to be. Now I'm wondering if anyone has had this issue, how I might go about debugging CodeProject to solve it? With these hard freezes/random lockups I don't have any good stacktrace to go from.
System.
Blue Iris 5.8.1.5
i7 8700K Processor @ stock speed with integrated graphics
Nvidia GTX 1060 3GB (Asus)
Motherboard: Gigabyte Z370 AORUS Gaming 5
32GB DDR4 at 2400mhz stock speed
Samsung 980 Evo SSD 512GB (plenty of free space)
Multiple other internal drives for various reasons
Windows 10 latest updates/edition
Nvidia latest graphics driver direct from Nvidia.com (clean)
Installed most recent network card (Intel ethernet) drivers (clean)
Installed most recent intel graphics and chipset drivers (clean)
Cameras (4x with PoE switch)
Amcrest 5MP Turret POE Camera, UltraHD Outdoor IP Camera POE with Mic/Audio, 5-Megapixel Security Surveillance Cameras, 98ft NightVision, 103° FOV, MicroSD (256GB), (IP5M-T1179EW-28MM)
Camera settings (double checked on every camera and tried multiple different settings)
Main Stream
H.264
Smart Code: Off
1920x1080 resolution (even though the cameras can do 4k)
25FPS
VBR
Quality: 4
Max Bit Rate: 4096
Sub Stream
H.264
704*480(D1) resolution
15FPS
CBR
Bit Rate: 512Kb/S
Frame Interval: 30
Within Blue Iris, I have tried setting up the cameras to decode with “Intel”, “Intel + VPP”, “Nvidia”. and CPU only with NO hardware acceleration, the system will lock up and crash regardless of which is chosen if I have AI enabled. With Intel + VPP, sometimes I might get lucky and it might go nearly a full day without crashing, but it still crashes every day.
It’s obvious that something is not playing nicely with Blue Iris/CodeProject integration, I thought maybe it was the camera streams/encoding, but they are all H.264 and fairly easy bitrate at 1080p…it should not have any issue with 4 cameras at 1080p.
Any suggestions? If not, do you have any other thoughts or settings you might suggest I change to try to achieve normal system stability? I would absolutely love to use the AI integration as it makes my alerts actually useful.
Thank you!