Custom model on the same port #318
Comments
VorlonCD
added a commit
that referenced
this issue
Apr 8, 2022
…d, Deepstack tab no longer says ERROR) #314 - New remote Telegram command: RESTARTAITOOL (or RESTART). This will restart AITOOL only. #319 - New remote Telegram command: SCREENSHOT. Sends a screenshot of the entire screen AITOOL is running on. #319 - Added SHUTDOWNCOMPUTER telegram command. It will shut down in 10 seconds. #298 - Dont force 'Detection API' to be enabled (Unless NO other modes including Custom are enabled. Previously, if Custom was enabled, it would still try to re-enable regular Detection). This is so people that only want to run their own custom models dont have to run normal detection consuming extra memory and cpu. #318 - Tweak how lat/long is read from blueiris registry for use with SUNRISE-SUNSET #296 - FaceStorage path now checks to see if the folder is accessible. If not, puts the path under the regular settings folder. (Like in the case AITOOL folder was moved) #316 - Rework facestorage manager. Still havent had time to make the UI for this working but all faces are still stored when found. See previous commit details for how it works. May have fixed a bug where the FACES.JSON file gets too large or corrupt. #316 - Only save every 30 seconds (or on demand) to reduce chance of settings file corruption (Controlled via 'SaveSettingsIntervalSeconds' in JSON settings file). prior to this, it was saving for EVERY image that was processed. - Pause now always defaults to all cameras rather than selected. - CTRL-MOUSEWEEL to change font size in History and Log - Added INNO setup files and automatic build of the installer (Before it was relying on Visual & Installer, a paid product) #301 - Some prep work with fonts and scaling for eventual move to .NET Framework 6. Changed default font to Segoe UI 8.25 for most controls. This can be set in the settings JSON file via "DefaultFont".
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Hi
Using AiTool 2.1.6.0 and would like to use the custom "general" model (from ipcamtalk) and the custom "Exdark" model.
I can run it manually on the same deepstack instance/port with this command line:
deepstack --MODELSTORE-DETECTION "C:\DeepStack\Custom" --PORT 88which then gives this:
However in AiTool it looks like I need to run each custom model on a different port and custom folder and not possible to disable "Detection API".
Which gives me 88 for detection api, 82 for general and 83 for dark.
Is it for performance reasons that is needs to be on different ports ? As in upload multiple files at the same time.
Maybe make it possible to disable "Detection API" when custom models is enabled ?
And then in "Edit Ai Server" I have this
as I don't want to use the inbuilt model from Deepstack.
And how do I control so "dark" model is only used when dark ?
Seems I can only edit times for Primary url, which should always be active.
Maybe I need to make "dark" the primary and "general" the linked one?
The text was updated successfully, but these errors were encountered: