Answer to 1 & 2 is yes. The answer to 3 is no. So I will try your suggestions.
I'm in US central time zone but I regularly collaborate with colleagues in Sweden Switzerland and Australia, so no worries about delays. Whoever is available whenever is fine. I just want to give you some feedback on things than might need to be added or clarified in the instructions. But overall, the installation and setup was pretty smooth, it probably helped that I had a P4B with "Buster" and an old version of node-red already setup. I'm also retired so I have no set schedule.
One of my early tests is on my main desktop I have the dashboard for my AI system running in a Chrome window and your dashboard from the Pi4B with the default four stream display in another Chrome window. My dashboard regularly goes dead with "Aw snap! something went wrong" and needs a manual refresh to start working again. It will be interesting to see if yours is free of the problem. It may very well be a node-red version issue, my system is v1.1.3 -- its been running a long time, "if it ain't broke, don't fix it!" and my issues seem to have started after some Ubuntu 16.04 updates. But if mine dies and yours doesn't I'll update to at least the v 1.3.1 that I have on my Pi4B.
I'll be attending a 50th reunion tomorrow so probably not much more will happen until Monday.
Edit:
One weird issue I just noticed is the Camera-2 mini display has gone blank but if I click it, Camera-2 displays in the main window.
Edit2:
Clicking the presets button brought back the Camera-2 small window display.
Thanks for your feedback.
As I suspected the directory/directories in all three storage areas weren't set-up.
I'll have to modify our sas-org.sh script to create these directories IF THEY DON'T EXIST.
The daily housekeeping routine, that runs at 5-minutes past midnight, deletes the 'day' directory and creates a new one, so once HSS is running all the housekeeping will work like magic. The daily houskeeping is our way of restricting the amount of disk space used otherwise HSS would gradually use more and more space. This means each day's recording is available for a month before that particular directory is re-cycled.
So from what you said...does that mean that you can now make recordings and play them back??
If try to fix some of the issues you mentioned in your feedback to improve our website's layout/rendering.
Have a great reunion.
EDIT 1:
I've modified the org-sas.sh script so it now creates all the 'day' folders (01 to 31).
So that will sort out that particular 'install-issue'.
EDIT 2:
If you wanted to you could re-run Step 1 and Step 2 on the HSS Setup SAS tab or just leave things as they are as the magic will happen at mid-night (just like Cinderella).
I tried again this morning snapshots and recordings now work. Really good job!
My comment about the dashboard was that it wasn't clear to me that I had to open the dashboard to see the HSS main display, with all the nginx installation and testing I mistakenly assumed I needed to view something on port 80. Doesn't that editor window button just open the dashboard "editor" panel?
I'd suggest clarifying the instructions:
Go to the tab-flow labelledā¦ HSS initialization and click the Inject node labelled āget the configuration fileā.
This action should result in some video streams appearing in the small monitors and the main display area.
and add "on the the node-red dashboard at: node-red_IP:1880/ui"
Basically I just followed the instructions on the website and other than this recording/snapshot directory issue, which you have now fixed, and a few user misunderstandings it worked out of the box. Having the public streams configuration available to verify things are working before jumping into setting up my own cameras is genius!
I look forward to diving in deeper when I return on Monday.
Side note, the "Camera-2" small display seems to drop out (happened multiple times) but clicking its blank window still brings it up in the main display window. Clicking the "Presets" button brings it back for a while. My Python rtsp threads have to deal with cameras going away and automatically restart when they return I loop trying to re-open the stream and sleep for five seconds or so before trying again. Not sure if or how node-red could do something similar.
It has been discovered the permissions on the 'snapshots' directory may not be correct which will prevent snapshots from being deleted. The 'housekeeping' script that handles this has now been updated.
So if you are affected by this issue, just go to HSS Setup SAS tab/flow and re-run Step 1 and 2.
I have no control over those cameras, so can't guarantee they will be available 24 x 7.
It will be interesting to hear if any other users are experiencing this.
EDIT:
It might be worth editing your config file so Camera-2's stream on Monitor 2 is on say Monitor 1 and see if the 'effect' moves or remains on Monitor 2. That might narrow it down a bit as to whether it is the source stream or something in our flow.
Been away for a couple of days, but I left the HHS dashboard (Pi4B) display running along with my dashboard (Ubuntu i7) and my dashboard was "Aw snap, something whet wrong" while the HSS dashboard Camera-1 main window display was still streaming. Looks like updating my i7 node-red version is added to my ToDo list. I'm impressed!
On what I've downloaded, with the initial Preset, Camera-3 and camera-4 appear to be duplicates of the same stream, which seems to drop out quite often and leaves a static frame in the small window. Camera-2 drops out often but always seems to leave a blank small window. All display in the main window when when I click the small window but the only way to restore the small window displays is to reload the Preset.
I am sure I will have more questions once I start trying my own cameras and some of the more advanced features.
Yes Camera3 and Camera4 are duplicates. I need to find another different one that has a good update frequency. Someone else has mentioned about Camera2 dropping out. So we may remove that stream.
I've just written some content for the FAQ part of the website explaining how to delete images.
Give it a read and let me know if it makes sense.
Unless you can find others as good as the Camera-1 stream, these are better than nothing, maybe just add a note in the setup instructions about the limitations of publicly available streams over which you have no control.
It really is a nice way to verify the installation and get familiar with how it works before diving into the specific details of customizing for specific needs.
After playing a bit with the snapshots and recordings I like the way differing resolutions appear to be automatically handled. I like that php widget you embedded to let the main window play the jpgs and videos, but I like it even better that the jpg and mp4 files are available in the hss_media directory so I can access them with whatever apps I prefer via sharing the directory.
I do a similar thing with a Telegram Bot - it sends an in-line keyboard to my phone asking me which camera I want to take a snapshot from. Then captures a snapshot and sends it to my Telegram acct. The whole thing runs outside of HSS and is completely independent of HSS - just uses the photo media directory.
Wow. I'm going to have to try this out. Just the other day, I sent an email to the developers of my paid security system (Sighthound Video) asking when they will ever support install on a raspberry pi. I need to update the computer it is running on.
Combine a surveillance system with an effective object detection system (having a low-enough false positive rate) with detection push notifications, and you have a very effective security system.
I hope to morph it into a front end for my AI security system, as this is so much better than the minimal node-red system I came up with.
I see you at least have some stubs for object detection built in, but since TensorFlow.js doesn't seem to support the Coral TPU or other AI co-processor devices, I doubt you can effectively run it on IOT-class hardware (Pi4B etc.), especially with multiple cameras.
The Coral TPU is ~$60 in the USB3 version and they have various MPCIe modules for ~$25 each.
Basically where your object detection pushes images into tf_coco_ssd I'd push them into my Python based AI subsystem.
