AI-on-the-edge-device
AI-on-the-edge-device copied to clipboard
jomjol
Outsource calculations to independent server
The Feature
I know this probably requires a rewrite of the whole system. But my idea is that the esp is only making the image and sending it to a server which calculates the values etc. The server could be another esp or a Raspberry Pi or even a server located in a datacenter. (the user has to provide the server aka you don't have to host them) I have the feature request because my systems are slow af and requires like 10 minutes to render the web interface. (freshly rebooted) to my conclusion the esp is over worked and doesn't have any spare power to render web GUI
I understand your feature request. It was requested before. But the whole point of this project is to have it as simple as possible and that fulfilled with keeping it inside a single device.
There have been discussions about this before.
are slow af and requires like 10 minutes to render the web interface.
It looks like this is the root cause of your request! And most likely something that an be fixed. Please open a discussion and provide some more infos about sour system (RSSI, logs of the startup, distance to your accesspoint, ... Also try it with a different SD card and search other discussions about performance issues.
It is still possible to keep it in a single device. Just host the server on localhost and set the rest to send the data to localhost instead of a other server.
As long as there are no major board or configuration issues the hardware is capable to drive all without major delay.
E.g. I have cycle times of around 15 sec and quite responsive UI.
Next improvement would be a board with modern ESP32S3 SOC. There you get cycle times between 5-10 sec.
It is still possible to keep it in a single device. Just host the server on localhost and set the rest to send the data to localhost instead of a other server.
This is how one would do it in general. This would mean to add several additional REST APIs. But we are short on memory and every additional REST API equires some extra memory!
are slow af and requires like 10 minutes to render the web interface.
It looks like this is the root cause of your request! And most likely something that an be fixed. Please open a discussion and provide some more infos about sour system (RSSI, logs of the startup, distance to your > accesspoint, ... Also try it with a different SD card and search other discussions about performance issues.
Did you at all at all consider to first work on the root causes?