CodeProject.AI Version 2.0

Wait a minute.... THERE'S CORAL SUPPORT NOW? (re: 2.1 changelog)

Does it only work on a pi? Well, I assume not because that would be silly. I've had a usb coral sitting in the drawer for months just waiting for this glorious day!

Please confirm if true
 
  • Like
Reactions: David L
Wait a minute.... THERE'S CORAL SUPPORT NOW? (re: 2.1 changelog)

Does it only work on a pi? Well, I assume not because that would be silly. I've had a usb coral sitting in the drawer for months just waiting for this glorious day!

Please confirm if true
Looking at the modulesettings.json it only works for the platforms highlighted in green. Since a Raspberry Pi using linux-arm64 OS it will work on the Raspberry Pi.

1680455505834.png
 
  • Like
Reactions: David L and tofu
Cat shadow = Bird
1680614423687.png
1680614711310.png
 
It would be great if there was some way of applying a vector type layer to the view and height of camera to allow perspective to be part of the processing so it would then have a view of basic object size based on it's position in the frame and could scale an object to a standard size ratio for image processing against the AI DB. Obviously the best solution would be depth sensing cameras using lidar or similar but think those are a way off from consumer level pricing for security cams.
 
That shadow is clearly from a giant snail. I guess there is no "giant snail" algorithms however.
What I really need to do is look for a better lighting source, maybe it will minimize the dark shadows. What I have is a LED bulb (40 watt brightness) in an outdoor wall light which does have diffuser glass. I have been wanting to add one on the other side of my garage too...
 
It would be great if there was some way of applying a vector type layer to the view and height of camera to allow perspective to be part of the processing so it would then have a view of basic object size based on it's position in the frame and could scale an object to a standard size ratio for image processing against the AI DB. Obviously the best solution would be depth sensing cameras using lidar or similar but think those are a way off from consumer level pricing for security cams.
Sound great but how big of a computer do we need and I think the cam would be a little out of our price range even if we buy it from Andy.
 
Interesting that Coral is now at least showing as to be potentially be supported. I'm actually testing Frigate on a Ubuntu Server (bare metal) with my newly received M.2 Coral but I guess I can use that same box to install a CodeprojectAI docker that can utilize the coral. Very interesting times.

I love my Blue Iris so I won't let go of it yet but been having issues with CodeProjectAI in Windows (rolled back to Deepstack for now). In addition to that, I've been using my GTX 1080 with this Blue Iris server for faster detection and I'm having an additional 40-50watts because of that. My bare-metal linux server is like 15 watts while the Coral is being active with frigate.

Really excited for this!
 
CodeProject has a lot more bells and whistles than Deepstack has and the program is much larger as well, thus needing more horsepower.

Couple that with a 4th gen CPU and GPU with only 1GB and you could run into issues.

I have an older laptop that I use as my test before I do major BI updates. It was the computer I was using all thru when Deepstack came out and I was testing BI versions the moment it came out. It ran Deepstack just fine (albeit times were slow in the low 1,000s) and it won't run CodeProject at all.

You need to start cutting out models and number of cameras and try it one camera at a time until you see it struggle.

Or go back to Deepstack. I am still on Deepstack as there seems to be way too many issues for many folks here for me to risk it on my BI computer.
d
Could you send me that thread on how to go back to DS from CPAI? DS worked soooooooooo much better for me. I pretty much done with CPAI for now.
 
Sound great but how big of a computer do we need and I think the cam would be a little out of our price range even if we buy it from Andy.
Agreed, was just thinking about some (very) future enhancements. Cost\Processing will be a massive constraint for some time.
 
  • Like
Reactions: David L
Now it looks like folks around here are saying DS is dead.
Dead in support, or dead in not working?

I switched from DS a few weeks back, which was worked good for me. I have an old 4th gen CPU and 970 Graphics card.
I ran the GPU version of DS, no problems, doing the same with CP.AI and it seems to be working fine too.
1680781360558.png

What problems are you having?
 
  • Like
Reactions: gwminor48
Dead in support, or dead in not working?

I switched from DS a few weeks back, which was worked good for me. I have an old 4th gen CPU and 970 Graphics card.
I ran the GPU version of DS, no problems, doing the same with CP.AI and it seems to be working fine too.
View attachment 159241

What problems are you having?
I‘m getting triggers associated with clouds, bugs, small animals, rain. Kinda like before DS. You know, like back when we just relied on how we tweaked the cameras. I updated CPAI and so forth. But for some reason I’m not even getting alerts recordings saved. I had this happen before and I can’t remember how I fixed It. But I’ll get a “nothing found trigger” on incidents I want. I think before it had something to do with the DB but just don’t remember. I probably should take notes. When I was using DS it worked so well I thought my camers all quit working. It would go days without getting any alerts. And when it did send an alert you could rest assure it was a person or car. It could be cloudy, rainy, etc and I got no false triggers. I had CPAI kinda working a coupe weeks ago, not as good as DS. But for some reason everything is acting flaky right now.

Oh, I read a tread on here a couple days ago that someone DS just died. Maybe the support for Window’s.
 
  • Like
Reactions: David L
CodeProject.AI version 2.1 changes

Release notes for 2.1
  1. All modules are now equal. On first startup we install Object Detection (Python and .NET) and Face Processing, but
    these can be uninstalled. The Server installer itself no longer installs these modules: they are auto-installed on
    first startup by the server itself.
  2. Reworking of the Python module SDK. Modules are new child classes, not aggregators of our module runner.
  3. Reworking of the modulesettings file to make it simpler and have less replication
  4. Improved logging: quantity, quality, filtering and better information
  5. Addition of 2 modules: ObjectDetectionTFLite for Object Detection on a on Raspberry Pi using Coral, and Cartoonise for some fun
  6. Improvements to half-precision support checks on CUDA cards
  7. Modules are now versioned and our module registry will now only show modules that fit your current server version.
  8. Various bug fixes
  9. Shared Python runtimes now in /runtimes. No more AnalysisLayer folder. These are all now in /modules
  10. Tested on CUDA 12

Is there going to be an X86 version of the docker with coral support pushed out soon? I run the docker on my UnRaid server that also has a Coral installed in it. Been waiting for this update to come out.
 
  • Like
Reactions: David L