The Typical picture of a Perp on Nextdoor-type Apps with Consumer Grade Cameras like Ring, Nest, Arlo, Canary, Wyze, etc.
- Thread starter wittaj
- Start date
Chokolinho
Getting comfortable
From what I've seen on trading platforms, many people buy Reolink cameras and protect them fiercely because they are so good.
Here is a new brand to me - LUMA - and of course Looney's favorite taking a video with cellphone of the monitor LOL.
Typical case of expecting a wide angle camera to ID at distance. Anyone recognize these door checkers?
Typical case of expecting a wide angle camera to ID at distance. Anyone recognize these door checkers?
Anybody recognize these two that broke into cars....and interesting that the camera captured the one of the left shadow but not his upper body LOL
The OP asked if anyone could clean these up to get a face and plate LOL
And then someone posted that AI came up with this:
And this LOL
And then someone posted that AI came up with this:
And this LOL
mat200
IPCT Contributor
- Jan 17, 2017
- 17,764
- 29,912
OpenAI gave me this analysis based on the first picture
Bottom Line
- Vehicle: White full-size pickup, likely late-1990s/early-2000s Chevy/GMC
- License Plate: Not readable from provided image
- Suspect: Present but not identifiable
- Evidentiary Value: Limited; suitable only for narrowing vehicle class, not positive identification
If you can provide:
- Additional frames, or
- The raw video file
I can perform frame stacking, motion deblur attempts, and plate-region isolation to push this further within forensic limits.
It would be interesting to try that out, no? Do you have a video that you could give it?
Or post how to ask that AI the question. I have no knowledge of using AI.
mat200
IPCT Contributor
- Jan 17, 2017
- 17,764
- 29,912
It would be cool to see if it can actually do this.
I only have the free account, so I suspect I would hit some sort of limit quickly on a test like this.
In general :
Don't trust the AI .. last time I tried to get it to do the real work it asked for, it said it would and did not do it.
Basically the LLM AI is really good with words and predictive generation of words, and bad with actually doing the work.
ChatGPT I tested, did a great job at promising me to do some work .. then after asking me all sorts of questions about the work ( even "confirm that you want me to do X" ), it stated "You have run out of time, do you want to pay me for more time" .. )
There are tons of different AI versions / companies / products out there. They are constantly getting better, and I am certain one of those will do a better job at attempting to analyze multiple images, I just have not yet found it.
Sounds like a lawyer. Book al sorts of 'time' to charge as much as possible but never get you anything.
Can anyone make out the plates or the vehicle make/model that took out a mailbox and completely missed a driveway
A little stop-n-rob at the stop-n-go. Why can't businesses get better cameras. And certainly the collage of images makes it easier to ID.
Certainly someone can make out the car model cruising down the street with the passenger out checking car doors.
Wyze don't the home owner have a cleaner capture of this pickup truck that sideswiped his car parked on the street?
