this post was submitted on 21 Aug 2023
250 points (100.0% liked)
the_dunk_tank
15914 readers
12 users here now
It's the dunk tank.
This is where you come to post big-brained hot takes by chuds, libs, or even fellow leftists, and tear them to itty-bitty pieces with precision dunkstrikes.
Rule 1: All posts must include links to the subject matter, and no identifying information should be redacted.
Rule 2: If your source is a reactionary website, please use archive.is instead of linking directly.
Rule 3: No sectarianism.
Rule 4: TERF/SWERFs Not Welcome
Rule 5: No ableism of any kind (that includes stuff like libt*rd)
Rule 6: Do not post fellow hexbears.
Rule 7: Do not individually target other instances' admins or moderators.
Rule 8: The subject of a post cannot be low hanging fruit, that is comments/posts made by a private person that have low amount of upvotes/likes/views. Comments/Posts made on other instances that are accessible from hexbear are an exception to this. Posts that do not meet this requirement can be posted to !shitreactionariessay@lemmygrad.ml
Rule 9: if you post ironic rage bait im going to make a personal visit to your house to make sure you never make this mistake again
founded 4 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
For what it's worth, this guy is likely lying about the actual capabilities of this system.
Tracking the time the person has spent in view is a legitimate task this software can accomplish (although this was called person/object detection/tracking long before AI was the newest buzzword).
The number of cups the baristas are making? Likely bullshit.
Exactly, tracking time in frame is a pretty standard CV task. Tracking # cups of coffee made would be require more sophistication than any cafe in the world could afford, if it's even possible to do reliably.
Purely guessing for fun:
If it can reliably recognise workers, and every worker is only making coffee for their own customers: count how many coffees the worker entered into the POS system based on timestamps.
Yeah that would be one way to do it without any computer vision tech. Recognizing workers might be a stretch even, that would require a model to be trained on shitloads of pictures of every single employee.
I think it can be pretty easily setup by the company contracting out this software with a single consultant/engineer
You can see in the photo that there's a clear division of customers and employees separated by the counter
So hardcode in the counter coordinates on the camera screen and any human detected to the right of the counter can automatically be assumed to be a worker and anybody left of it can be assumed to be a customer
From how it looks in the screenshot, every worker has a differently colored tracer following their movement. Why mess around with face recognition, when you can just pin an led badge to every worker
Because you can call it AI and squeeze an extra comma out of the coked-out VC ghouls funding your company
You could still do that without using any advanced AI. It's not like the customers or VC ghouls are going to know the difference.
Calling algorithms AI is a time honored tradition.
Took the words out of my mouth
I do some applied ML on photos/videos and I feel like this should be pretty simple
Manually map out the coordinates of the counter and add a bounding box of coffee cups/mugs in addition to your baristas. Count how many times an overlapping coffee mug and barista bounding box enters the bounding box of the counter
I guess that would work if you structured the shop's entire workflow around being recognizable by the program. Even then, pairing this with employee recognition and considering all the edge cases it would be very hard to pull off. It would be a really cool problem to hash out if it wasn't for such a cartoonish evil application.
Its crazy that this would be a thing AI would need to do when any half descent accountant could simply measure cups-poured by shift and sift the data for the most productive employees. Even then, the bottleneck in this shop is certainly not the staffers themselves. The work space is tiny and the equipment is antiquated. You're investing god knows how much in AI when you'd be far better off renovating.
You could just track cups via the tills. I don't get it, none of this adds anything at all that doesn't already exist. You already produce a record of transactions per employee via receipts produced in the cash registers.
The last decade of western technological advancement in a nutshell.
The tracking doesn't even matter, the number of cups is not even a useful statistic. Why would anyone want to track that?
I assume the 5 or so people behind the counter have their own responsibilities and work as a team instead of competing to serve the most cups.
WOuldnt the number of cups be tracked by sales anyways?
99% of applications for neural networks are just replacing sensors and measurement systems with a crude approximation implemented using image recognition that works (at best) only in the ideal demo scenario they present to investors/clients
A friend of mine worked for a company that was trying to sell a warning system for unfastened seatbelts using a webcam in the console