Version 6.0.1

[hxhoang]

I have Proxmox, Windows 10, GeForce 1650, built-in AI, BI crashes with this error

My Proxmox Windows 11, RTX A2000 setup works fine with built-in AI in Blue Iris 6.0.1.9.

Replies from Gemini with your error message:

The DmlExec exception in Blue Iris 6.0.1.8 indicates a failure within the DirectML (DML) execution provider. This typically happens because the AI runtime cannot properly communicate with your GPU drivers or is missing essential Windows system components required for hardware acceleration.

1. Update Visual C++ Redistributable
This is the most common fix for Blue Iris v6 crashes and AI initialization errors. Version 6 requires the latest v145 libraries.

Download and install the latest Visual C++ Redistributable (x64).

Reboot your computer after the installation to ensure the new libraries are registered.

2. Update GPU Drivers
DirectML (DML) relies on your graphics card's DirectX 12 capabilities. Outdated drivers often lack the specific instructions needed by newer ONNX models.

Intel Users: If you are using Integrated Graphics (like Iris Xe), download the latest drivers directly from the Intel Support site.

NVIDIA/AMD Users: Ensure you are on the latest "Game Ready" or "Studio" drivers.

3. Clear the AI Model Cache
If a previous version of Blue Iris or CodeProject.AI left behind corrupted temporary files, initialization may fail.

Stop the Blue Iris service and the CodeProject.AI service.

Navigate to your AI installation folder (usually C:\Program Files\CodeProject\AI).

Look for a temp or cache folder within the ONNX or specific model directories and delete its contents.

Restart the services.

4. Perform a "Full" Reinstall of Blue Iris
For version 6.0, the developer recommends a fresh installation rather than just an in-app update to ensure all new directory structures and dependencies are correctly placed.

Backup your settings first (Export in Blue Iris settings).

Uninstall Blue Iris, then download the latest full installer and reinstall. Your registry settings and camera configurations are typically preserved during this process.

5. Disable GPU Acceleration (Troubleshooting Step)
If you need your cameras back online immediately and hardware acceleration is failing, you can force the AI to use the CPU.

In the Blue Iris AI settings, change the execution provider from "Default" or "DirectML" to "CPU". If the error disappears, the issue is confirmed to be with your GPU driver or the DirectML compatibility layer.

 

[barsik89]

My Proxmox Windows 11, RTX A2000 setup works fine with built-in AI in Blue Iris 6.0.1.9.

Replies from Gemini with your error message:

The DmlExec exception in Blue Iris 6.0.1.8 indicates a failure within the DirectML (DML) execution provider. This typically happens because the AI runtime cannot properly communicate with your GPU drivers or is missing essential Windows system components required for hardware acceleration.

1. Update Visual C++ Redistributable
This is the most common fix for Blue Iris v6 crashes and AI initialization errors. Version 6 requires the latest v145 libraries.

Download and install the latest Visual C++ Redistributable (x64).

Reboot your computer after the installation to ensure the new libraries are registered.

2. Update GPU Drivers
DirectML (DML) relies on your graphics card's DirectX 12 capabilities. Outdated drivers often lack the specific instructions needed by newer ONNX models.

Intel Users: If you are using Integrated Graphics (like Iris Xe), download the latest drivers directly from the Intel Support site.

NVIDIA/AMD Users: Ensure you are on the latest "Game Ready" or "Studio" drivers.

3. Clear the AI Model Cache
If a previous version of Blue Iris or CodeProject.AI left behind corrupted temporary files, initialization may fail.

Stop the Blue Iris service and the CodeProject.AI service.

Navigate to your AI installation folder (usually C:\Program Files\CodeProject\AI).

Look for a temp or cache folder within the ONNX or specific model directories and delete its contents.

Restart the services.

4. Perform a "Full" Reinstall of Blue Iris
For version 6.0, the developer recommends a fresh installation rather than just an in-app update to ensure all new directory structures and dependencies are correctly placed.

Backup your settings first (Export in Blue Iris settings).

Uninstall Blue Iris, then download the latest full installer and reinstall. Your registry settings and camera configurations are typically preserved during this process.

5. Disable GPU Acceleration (Troubleshooting Step)
If you need your cameras back online immediately and hardware acceleration is failing, you can force the AI to use the CPU.

In the Blue Iris AI settings, change the execution provider from "Default" or "DirectML" to "CPU". If the error disappears, the issue is confirmed to be with your GPU driver or the DirectML compatibility layer.

Thank you. I did all this, and it helped. BI works without errors. But what I noticed is that when AI is triggered, my CPU load is 100%, and my GPU load is 15-20%. As a result, processing AI triggers takes 3-5 seconds, and everything slows down. I allocated 10 cores in Proxmox and updated to the latest version of BI. Nothing changed.

 

[hxhoang]

Thank you. I did all this, and it helped. BI works without errors. But what I noticed is that when AI is triggered, my CPU load is 100%, and my GPU load is 15-20%. As a result, processing AI triggers takes 3-5 seconds, and everything slows down. I allocated 10 cores in Proxmox and updated to the latest version of BI. Nothing changed.

Looks like CPU still carries the load instead of GPU.
It is quite tricky to run BI in VM. I have 12 cores (i9-13900HX) allocated for my VM and CPU load rarely goes over 30%. Someone mentioned to lower "max connects" in main AI tab to 2-8 helped CPU load.
BTW, Do you see your GPU listed in main AI tab? Please note that BI normally crashed for the first trigger after SW restart. It should be normal for the following triggers.

 

[barsik89]

Looks like CPU still carries the load instead of GPU.
It is quite tricky to run BI in VM. I have 12 cores (i9-13900HX) allocated for my VM and CPU load rarely goes over 30%. Someone mentioned to lower "max connects" in main AI tab to 2-8 helped CPU load.
BTW, Do you see your GPU listed in main AI tab? Please note that BI normally crashed for the first trigger after SW restart. It should be normal for the following triggers.

My GeForce 1650 is selected in BI. I tried reducing the maximum connections to 3. It got better. But when multiple cameras are triggered, the CPU load reaches 100%. Before, I used Blue Onyx on the GPU, and it works well. I guess I'll have to keep using it.

 

[CAL7]

I'm trying to put in a make time of 0.5 seconds and a break time of 4.0 seconds. In the motion sensor page, I clearly see the make time parameter, but I do not see an explicit break time parameter. There is an option after make time of how many times within how many seconds. Is that the equivalent of break time, and how would I use that to get a 4.0 second break time? Thank you!

P.S. ChatGPT says I'm looking for a reset parameter. But I don't see that either.

 

[bp2008]

Break Time is not in the motion sensor configuration panel, it is out in the main Camera Settings > Motion/Trigger tab.

 

[hxhoang]

Has any one upgraded BI to 6.0.1.10? It seems to force AI to all cameras and AI cannot be disabled on camera level. It is OK on previous versions.
BTW, For 6.0.1.9, I got error message: Failed to load class names from: C:\BlueIris\AI\models\coco.names if I used "Trigger now" on camera to test AI eventhough normal triggers are still OK. Is it normal?

 

[Bruce_H]

BI Version 6.0.1.11 was just released to fix the AI enable/disable issue!

 

[Vettester]

BI Version 6.0.1.11 was just released to fix the AI enable/disable issue!

LPR is not working in 6.0.1.11

 

[hxhoang]

My GeForce 1650 is selected in BI. I tried reducing the maximum connections to 3. It got better. But when multiple cameras are triggered, the CPU load reaches 100%. Before, I used Blue Onyx on the GPU, and it works well. I guess I'll have to keep using it.

Have you ever tried frigate? It works very well for me with iGPU (LPR) and pcie EdgeTPU (object detection). I have both BI on VM and Frigate on lxc running concurrently (mix of 8x 2k-4K camera) on the same machine. Performance-wise, frigate seems to be better on my setup.

 

[CCTVCam]

You need a better GPU, there's not much grunt on a little card like the RTX 5090, also only having 32GB GRAM on it will limit things.

32GB is a gold mine. Sell the card, get rich and go intergrated graphics. There's rumours of the 5090 becoming a $5,000 card due to memory price rises.

I'm not sure we'll see many more BI 5090 builds, it will be cheaper to le them rob your house...

I'm a grinch. I haven't put up Christmas lights in years!

I didn't even put up a tree. Been wanting $40 even for a small one. As I live alone, not worth it. So stuff that. I bought a pot of Hyacinths instead and went for a ofactory Christmas instead of visual. Worked well, they flowered on Christmas Day!

 

[CCTVCam]

Couple of querstions as I've yet to make the jump to 6.0.

1. Can you jump straight to 6.0.0.9 or do you have to install 6.0 1st?

2. Do you still need to uninstall your 5x build 1st?

3. Do you need to back up settings if uninstalling 5x and then reinstall?

4. Can you choose to use either YOLO 8 or 11 with intergrated graphics? I like the idea of using a much later model if the intergrated graphics can handle it rather than having to go back to 4 or 5. I'm on Intel UHD 630 graphics. I believe the intergrated chip has 184 shader units which I believe is the cuda core equivalent if that gives any sueful performance metric for those in the know on YOLO performance.

 

[CAL7]

1. Can you jump straight to 6.0.0.9 or do you have to install 6.0 1st?

That's what I did and see no problems.

2. Do you still need to uninstall your 5x build 1st?

I believe the strong recommendation is to NOT uninstall BI5 until after BI6 is running. Again, that's what I did (in fact, BI5 is still on my system).

3. Do you need to back up settings if uninstalling 5x and then reinstall?

I didn't uninstall, so a moot point for me. But I did backup all my settings but haven't needed it.

 

[Fitz2380]

I have installed version 6.0.1.9 on a new computer. I am coming from a 5.8.x version that was using Deepstack for the AI and this was all running on a different computer. I installed the new version of Blue Iris on the new computer and imported the Blue Iris.reg file from the old system. All of my cameras are working, but I cannot get the built in AI to work. I have tried using the built in NVIDIA card and No GPU when selecting the Built-in AI but there is no difference. When there is a trigger from a camera I get the following message in the log file: Alert canceled [nothing found] 0ms. This does not seem like it is truly being sent to the AI for processing. I must have something setup incorrectly or is there some issue from bringing over my backup setting from the old computer. I did uninstall and reinstall and that did not resolve anything. Anyone have any ideas to try?

 

[Fitz2380]

I have installed version 6.0.1.9 on a new computer. I am coming from a 5.8.x version that was using Deepstack for the AI and this was all running on a different computer. I installed the new version of Blue Iris on the new computer and imported the Blue Iris.reg file from the old system. All of my cameras are working, but I cannot get the built in AI to work. I have tried using the built in NVIDIA card and No GPU when selecting the Built-in AI but there is no difference. When there is a trigger from a camera I get the following message in the log file: Alert canceled [nothing found] 0ms. This does not seem like it is truly being sent to the AI for processing. I must have something setup incorrectly or is there some issue from bringing over my backup setting from the old computer. I did uninstall and reinstall and that did not resolve anything. Anyone have any ideas to try?

Doh: There is a check box on the AI window for camera settings. "Primary object detection".... Make sure this is selected.. AI is now working for my cameras again.