You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am wondering, if we would have an efford, if it would be possible to run this software - basically the AI detection of those images (of my meter) - somewhere else (x86_64 machine)?
In my poor understanding, it should be possible to compile the code also for other architectures and I guess it will need a runtime environment. But I think, we would have the following benefits:
1.: More (CPU)-Power - avaibility to analyze even more pictures, to e.g. don't miss the switch between a "4" and a "5" - the last bit of the "4" could get recognized as a "7" - but If the AI Model (or algorithm) knows/sees lets say 1-2fps, it should see: "ahh there was a "4" only 5 frames ago, this cannot be a "7", this must be the last bit of the "4", so I report "4.8"....
2.: More cameras to use (e.g. some simple RTSP Wifi-cameras, which will bring a much better picture, than the ESP Cam)
3.: Due to a higher (more often) reading, we can have more acurate reporting of the (water-)usage
I was just thinking, don't know if thats possible... Thats, why I ask here :) I know the initial approach was docker, but maybe times changed and it's again time to ask :)
I doubt whether a "normal" CPU is better suited for this; in the evaluation, the CPU power is not that important if it is not built for these tasks. The ESP32 takes me around 50 seconds for one round, but the ESP32-S3 takes around 20 seconds and that with the same CPU speed (160Mhz).
Somewhere on the Internet I once saw that a provider offered a USB-ESP32-S3 to accelerate the AI functionality of a "normal" PC (the PC sends the task to the ESP32-S3, the ESP32-S3 calculates and sends the result back to the PC).
well I mean a normal i71150 like its in my Proxmox machine... I think - and I know AI stuff isn't that good on a "normal" CPU, that a machine like this should handle more AI operations than an ESP32 ?!?
At least when using the Coral Stick as AI enhancement...?!
Sorry, but maybe I missunderstood something?! :) just want to get sure :D
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
Hi guys,
I am wondering, if we would have an efford, if it would be possible to run this software - basically the AI detection of those images (of my meter) - somewhere else (x86_64 machine)?
In my poor understanding, it should be possible to compile the code also for other architectures and I guess it will need a runtime environment. But I think, we would have the following benefits:
1.: More (CPU)-Power - avaibility to analyze even more pictures, to e.g. don't miss the switch between a "4" and a "5" - the last bit of the "4" could get recognized as a "7" - but If the AI Model (or algorithm) knows/sees lets say 1-2fps, it should see: "ahh there was a "4" only 5 frames ago, this cannot be a "7", this must be the last bit of the "4", so I report "4.8"....
2.: More cameras to use (e.g. some simple RTSP Wifi-cameras, which will bring a much better picture, than the ESP Cam)
3.: Due to a higher (more often) reading, we can have more acurate reporting of the (water-)usage
I was just thinking, don't know if thats possible... Thats, why I ask here :) I know the initial approach was docker, but maybe times changed and it's again time to ask :)
BR, Marcel
Beta Was this translation helpful? Give feedback.
All reactions