Camera pause is not working for the Asahi example stream.

If the user switches to a different model on dashboard, the model must be reloaded in check_model. This is not the case right now...

...if the uplink is slow. This needs to become compressed. Maybe with h264 and jmuxer?

Backup of Database and Data directory

...for Raspi and Nvidia GForce.

train_worker_gpu.py: sql_sequence should buffer jps in the right size, on harddisk and in memory.

This does not work with BeQuiet server

...would easily allow to switch external things, link sound and vision.

They should. Probably the best place to put scaling is the append-function in c_tfworkers.py. Then it could be removed from the two other positions.

Speed up and lock display while screen is renewed,,,

Users currently experience moments of uncertainty during data processing or content loading due to the lack of visual feedback. To improve user experience, I propose adding a loading indicator (e.g., spinner or progress wheel) that signals to users that the system is actively working. The loading indicator should be universally applied across the app during any significant loading or processing times.

...like the event frame directory to speed up access.

...there might be a loop too tight somewhere in the l_buffer system. We need to check...

SRTMP, TLS for RTMP

Option to hide movements, that are not part of the selected conditions. Target: Keep focus on objects of interest in lively environments.

https://coral.ai/products/accelerator

The current User Interface is pretty ugly. The plan is to make it more Userfriendly, but keep the functionality.

make the parameters for the training (like image augmentation and others editable) in the school dashboard

https://www.tensorflow.org/lite/

...should be more safe because the environment is outside the source dir.

...needs sorting out. For example:
Make one modulwe out of onecam.html, onedetector.html, oneeventer.html?
Integrate c_canvas.html into this and into index.html?

...using Tensorflow Data library.

Now its 3.0.5

... though they are only needed in very small versions:
/schools/views.py

There already is a feature to set conditions for starting an alarm, but the plan is to dedicate an entire new menu point to this.

... 10 seconds each. And do not join them to event-videos.

https://stackoverflow.com/questions/16658873/how-to-minimize-the-delay-in-a-live-streaming-with-ffmpeg
This seems to work nicely: ffplay -fflags nobuffer -flags low_delay "rtmp://192.168.125.56/bcs/channel0_main.bcs?channel=0&stream=1&user=user&password=password"

...very often the video in the emails does not show.

after the next screen is build up.

every 10 minutes purge MP4 C0001_2022... that are older than 10 minutes.

Reactivate Redis-mode to make vars process-global.
Put defaults into the table.
Signal on write to table --> refresh Redis-var.

...use release instead.

Check levelone FCS 9364
https://www.level1.com/level1_de/onesecure-videouberwachungs-management-software-64-kanal-57105407

The change to consistent use of aio has made some of the imports redundant. We need to clean out these declarations

We need an indicator for disc space and an automatic cleaner, that starts deleting when disk space gets tight.

... right now we have two, one in passwords.py and one in the database.

Delete events and by number, by age
Same for trainframes

Like Reolink E1 Zoom...

...are still missing.

...with Standard Keras Things. For Example: BinaryAccuracy for cmetrics...

getbmp, getbigbmp etc. plus the screens that require no login.
We need to introduce security tokens for these...

...in the PTZ control.

I think not. Also: the show/use checkboxes on the Cam page should be active when the detector mask is being edited.

On the Raspi-System getting predictions from the predictions-server: If the tag information of an image gets changed after the initial sending, the tags on the predictions-server is not updated. The next training might be faulty.

(seen in Kalymnos)