I really like that we have camera feeds on the site (https://dallasmakerspace.org/cam/). Is it possible to add a 15-30 minute timelapse to see exactly how much activity is going on?
Are there any concerns around privacy / resource usage etc?
I coded up a quick prototype to show what I mean: https://test.mandar.dev/dms-cam-timelapse/index.html
Probably. The publicly accessible camera feeds are intentionally a lower resolution so it’s easy to tell if a resource is being used but not necessarily by who or what their license plate is.
We can keep the same low resolution, but with a short buffer.
I think I get what you mean. If making a video that’s encoded so a browser could play it then it would be pretty resource intensive to continuously make those new videos. Another option could be having the browser load like 10 images per camera and use client side script to flip through the images so it looks like a video but isn’t (and thus doesn’t need to be encoded server side as a video).
Encoding on server is not expensive at all. I am running the prototype on a tiny 1 core VM and it has barely any CPU utilization.
The individual images are ~25kb and the video is ~250kb. I am stitching together 30 images.
Why is watching people on a camera feed important to you?
Is there a real use case that will increase the feeds’s value to the membership?
We already have this general capability at the space via the camera servers. It is access controlled for obvious privacy reasons.
Appears you’ve already built what you want.
It would give a better indication of how busy an area is over an arbitrary amount of time like 10 minutes compared to the information conveyed by a single image. It’s not a bad idea. Dunno if the juice is worth the squeeze but it’s a neat idea.
In your opinion, what would be an appropriate sample rate? Single frame per minute? Isn’t that about our current capture rate?
A sample rate of 30 minutes apart if you are trying to measure area usage that would seem more appropriate. People that come in to do work rarely do less than that. My SWAG would be 90 minutes minimum time spent at DMS on average.
If goal is to sample how long people stay at DMS, then 10 minutes would be more reasonable to get an better assessment - but that assumes you are willing to track that person around.
RFID door rate will give you some idea of arrival (but this involves Personally Identifiable Information), but then there’s tailgating, walking up ramp when door open.
But no telling when they leave.
Also define “activity”, are two people sitting and talking an activity? Is it just BS’ing or it it a discussion about a project or a discussion about DMS? I was at DMS the other night, sitting with 2 other folks for over an hour. During that time, a third person came to collect a donation I promised for DMS (which is why I came), with the other two we discussed equipment status, new airline inline instillation, training, other “BS” items, answering several other members’ questions they came over to ask. I consider all of that “activity”, even the social part (which builds community cohesion).
I think what your doing is an interesting study and useful information. But before capturing the information, define how you will interpret it and what you want out as output. That will control the input.
1-2 frames per minute would make sense to me. I’m not sure what we do now, but having the historical data (say 5-10 minutes) would avoid some confusion when a user steps away from a tool for a few seconds at the moment the image updates, which is an issue I’ve run into on the lasers before.
I agree with Luke here, I like the idea, but it may be such a small use case it’s not worth the volunteer time to implement.
One use-case is to see if the 3D printers are in use. It’s almost impossible to tell from a single frame. But really easy from the timelapse.
I have the timelapse working for myself. If this is something that will be useful to others then I am happy to put in some more time to make it more robust and update the dms/cam webpage.
But just like any other resource, there is the chance that between the time you look at it and the time you show up someone else could have jumped onto it … or off of it. There’s no way to tell from a camera when that person will be done.
I suppose you could work with the 3D Fab chair @engpin and see if there’s a way you could write a script or something to poll all the Octoprint jobs and report their status (and remaining print time). Since everything is volunteer-done, I wouldn’t ask someone else to do it, but obviously you have software skills. I would guess that if this were achievable that a lot of people would use it.
We already working on more involved status page that will show if tools are in use/last use time / etc, especially around the 3D Fab printers.
I do kind of like the idea of a time lapse, but at the same time we are currently trying to get the rest of infra cleaned up, so it wouldn’t be something I think we should add onto our plates for at least a little while.
I believe the camera snapshots that are available on the cam page are taken once every minute or once every 5 minutes. I can appreciate why we might want a time lapse on them for that reason - that being said, I’m a little uncomfortable with the fact it would make it easier to follow people’s movement over a previous 10-30 min period just via camera. (Like sure it is already doable, as shown by this proof of concept, however it actually requires effort)
That sounds interesting! How can I contribute? Is there a wiki page or jira task I can pick up?
I like this idea. Sometimes I like to know if the quantum duplexer is being occupied, and if so, is it just one person, is it one person processing a mile-high stack of unobtanium, or is it more of a spontaneous social gathering.
Ultimately I would process the footage (something akin to motion blur), generate a heatmap, identify trends and incorporate a prediction algorithm. Then I would buy an eight ball, hollow out the inside, machine a reaction wheel, and then I would have a magic 8 ball that tells me whether to go to the makerspace.
If you refresh your browser, you will see the cameras in one minute increments.
