Neural network server

In the current implementation, a neural network server is used to:
1) Detection of photographing the computer screen by smartphone: connection to the neural network server (Linux/Windows, version from 1.00) goes through the server of the complex.
2) Speech to text translation: connection to the neural network server (Linux support only, version from 2.01) goes directly from client machines.

Common description
The server should be downloaded from downloads page
It can be installed on local, or remote machine.
Supported OS: Ubuntu 18+, Windows 10+/2016+ (only 64 bit)
Data exchange with the neural network server is carried out using secure https-protocol.
The port for communication through a colon must be specified, because by default the neural network server is configured to port 27524, and if no port is specified after a colon, then 443 will be used.
To change port in Linux edit next file /etc/stnnserver/config and then restart service: sudo service stnnserver restart
To change port in Windows edit next file C:\ProgramData\stnnserver\config and then restart service STNN Server

Detection of photographing the computer screen by smartphone
Client settings for this task are configured here.
Also, for detection, you need to have on the server machine (where the neural network server will be installed) video-card GPU with CUDA support. Also NVidia drivers must be installed for this card.
Current implementation uses CUDA 12.0, this version covers both old and newer video cards. Here https://en.wikipedia.org/wiki/CUDA#GPUs_supported you can find supported models. First column "Compute capability (version)" is from range 5.0 to 9.0.
More CUDA cores and more powerful GPU will result faster recognition cycles.
Estimate usage of VRAM - minimum 2GB, and RAM - near 2.5GB, CPU loading is minimal.

Speech to text translation
Client settings for this task are configured here.
GPU with CUDA is not required!
Attention! The connection test checks only the availability of the server, and not the voice recognition functionality itself!

Possible errors while testing the connection (common)
Timeout or error connecting to the complex server - it is impossible to connect to the server of the complex (the server of the complex is specified incorrectly, invalid port, Firewall blocking or the server is not running);
Server response timeout - failed to receive a response from the complex server within the maximum allotted time;
Config has not been read yet - the server of the complex did not read the settings from the database;
HTTP Error XXX - most likely the "server:port" of the neural network is incorrectly specified or the neural network server is not running;
Network error 12007/11001 - the DNS name of the neural network server is incorrect;
Network error 12029/10061 - error connecting to the neural network server (the neural network server is incorrectly specified, invalid port, the Firewall blocking or the neural network server is not running);
Network error 12002 - timeout for processing a request by the neural network server (large load);
Network error 12044 - client certificate error when connecting to the neural network server.

Possible errors while testing the connection (only the detection of photographing computer screen with smartphone - Objects detection)
Runtime exception: The detector is not initialized - no graphics card with CUDA support or acc. drivers not installed.