MikeLud1
IPCT Contributor
You can also create a folder on your BI server, in this folder create empty text files with the name of the custom models you are using, for example ipcam-general.pt. Then point BI to this folder.
5.5.8 - June 13, 2022 - Code Project’s SenseAI Version 1 - See V2 here https://ipcamtalk.com/threads/codeproject-ai-version-2-0.68030/
- Thread starter fenderman
- Start date
gwithers
Getting the hang of it
- May 18, 2016
- 49
- 38
You can map a local folder on the host running the AI server Docker container to the AI server custom model folder. You can do it by starting the docker with parameters and that will map a local folder from the host to the AI server custom folder running inside the container. You can then put the custom models in the local folder on the host that will then be seen by AI server that is running inside the docker container.
Code:
docker run --restart unless-stopped -p 32168:32168 --name CodeProject.AI-Server -d -v /usr/share/CodeProject/AI:/usr/share/CodeProject/AI --mount type=bind,source=/some_local_folder,target=/app/AnalysisLayer/ObjectDetectionYolo/custom-models,readonly
Last edited:
I tried this. I installed the IPcam-dark and IPcam-combined in a folder on BI server. Pointed BI to it and it didn't work.
I will have to give that a try. I'm new to the whole Docker thing. Also, I know that Unraid handles dockers a little bit differently than a standard Linux operating system. That gives me some direction to start with though thank you!
biggen
Known around here
- May 6, 2018
- 2,688
- 3,060
I wonder why BI needs to know the custom model. Doesn't it just ship off the motion event to the CP AI server? I guess I don't understand why it needs to know what model is being used when the CP AI server already has this information. It seems like it frivolous information as far as BI is concerned.
MikeLud1
IPCT Contributor
When BI send the image to CP.AI it also send the model name to use, before doing this BI needs to know the available models.
biggen
Known around here
- May 6, 2018
- 2,688
- 3,060
I was thinking the model to use was specified in the CP AI dashboard as well. I must be remembered wrong as its been a minute since I set it up.
Ok so according to the log on the startup of the Docker, the files are contained at /usr/share/CodeProject. When I navigate to that folder /usr/share the CodeProject folder is not there.
edit Ran a file search on / for codeproject. No results. I'm totally confused. May just have to give up on running it in Unraid and run it locally.
edit Ran a file search on / for codeproject. No results. I'm totally confused. May just have to give up on running it in Unraid and run it locally.
Last edited:
I was experiencing the same exact issues as you. After uninstalling CodeProject.AI and reinstalling I am now see GPU for Object Detection (YOLO) instead of just Portait Filter.
MikeLud1
IPCT Contributor
For anyone that has a dedicated LPR camera can you please PM me 2 day image and 2 night images so I can test how the ALPR module I am working on works for other cameras then mine.
gwithers
Getting the hang of it
- May 18, 2016
- 49
- 38
That is because the folder location for CodeProject is INSIDE the docker container and when you go to the host running the docker you are actually searching the host. This is a tricky concept to get straight that is part of how containers work and not related to CodeProject. The way I found to facilitate this is to map a local folder "into" the docker container as per the parameters I posted previously. Basically it tells the containers to look at a specific host folder when referencing a location in the container. So if you put custom models in the host folder, it is also going into the container to that specific assigned location (i.e. the custom model location for AI Server). It is how I added a non standard model (delivery.pt) to AI server running in their docker. Any custom models you put in the mapped local folder on the host will show up in the AI Server custom models (see delivery model example below). It is no doubt a tricky concept to get straight. If this is still unclear, a Windows install is certainly more "natural" to interact with.
Last edited:
OfficerD00fy
Young grasshopper
You mean in the CP.ai server? Yes, just like the screenshot shows.
My problem seems to have resolved itself though. No idea why. It's probably just waiting to rear its ugly head again at an arbitrary date in the near future.
mk1shitrocket
n3wb
- Dec 23, 2020
- 7
- 3
mk1shitrocket
n3wb
- Dec 23, 2020
- 7
- 3
Does it run every available model even though I only have two(ipcam-combined,ipcam-animal) set to this camera?
MikeLud1
IPCT Contributor
Post a screenshot of your camera's AI settings. What look to be happening is the camera is using more then one model and you are seeing results for each model.
mk1shitrocket
n3wb
- Dec 23, 2020
- 7
- 3
The analysis post fact is different than real time, at least it was with DS and it would run all the models you had regardless of which ones you were using.
