Am I pushing technology too far?

[kimblebee]

How can it have a 360° view if it’s on the backpack? The only way to have an unobstructed view is to mount a camera on his head lol

 

[cobright]

Just running through a few more comments... and please believe I appreciate the criticism. I'm not arguing against any negative view here, god knows I kinda hate this thing myself but there might be a few misconceptions still...

This device simply records anyone and everyone anywhere without them being made aware. And is not limited to private property. I have an issue of me being surveyed and filmed, that data stored of me for ex , simply walking down the street in passing this child.

The device doesn't film anyone. It does take still photos of people it sees.

You don't seem trained on the psychology into potential sex offenders or kidnappers, you aren't working (as far as I know) with the proper authorities on development (again with training), etc.

I'm not trained is psychology or criminal justice. I am trained in analyzing human movement. In particular I've been using human movement models to improve routing algorithms in self-driving vehicles. The science encompasses determining intent and predicting behavior based on human movement. Things like picking which person in a moving crowd will change direction. So I don't claim any expertise in how a criminal might behave, but I do very well writing code that can determine how most people in a situation are behaving and then pick out the outliers.

Yes your pushing technology to far. Who wants to watch 8 hours of video from a child's backpack camera every night? That is not living.

I should say, the point of this little computer is that no one will have to watch 8 hours of video. There isn't 8 hours of video recorded anywhere. There is an initial set up period of a couple weeks where a proof-sheet of photos of people whom the computer found suspicious is presented for review each day. Suspicious generally meaning the person was watching the backpack kid more than other people present. The set up period lets the user or parent go through those people's photos and tag them as known-safe, or at least that the person has a reason to be watching the kid.

- Going through the footage every night can become a paranoia-reinforcer -- both for the parents and the kid. It means that -- to some extent -- the family will relive that horrible experience every single night. It's hard to "move on" and get past something if you spend time reliving it every night.
- Going through the footage every night can become such a chore that they just give it up, and -- if so -- what was the point?

Again, not really a chore. The first day I wore the backpack in testing I had 15 or so pictures to look through. It stayed about the same for a week then tapered off drastically. Now I don't get any assessment alerts most days when I take it out, unless I'm wearing a particularly funny t-shirt or if I trip and fall or something.

Continuing this thought, will it stand up to the kid shoving the backpack into a locker? Dropping the bookbag? Will it withstand a spilled drink?

The hardware is very robust. I don't have any interest in getting it rated waterproof, the backpack and its enclosure make it very splash proof. If it eventually takes a beating that does it in ... well off the shelf the hardware is under $200.

We're talking about a lot of effort for an event that's going to happen to a tiny number of people.

Well, a child-stranger abduction is very rare, only a very few hundred a year. But 200,000 children are abducted by a non-custodial parent each year, and over 1.5 million restraining orders are issued every year. Point being, while I'm developing this with a specific case-use in mind, its potential usefulness could be much broader.

Fallacy of forced choice. Any number of other options exist.

I presented no false dilemma. The options I listed are the only ones likely available based on the mother's stated intentions. My question relates to what action of mine is in the child's best interest considering these two possible outcomes. The mother has plenty of other options available to her but I believe her to be sincere in her stated intentions that without some form of improved surveillance she will sequester her child throughout the next school year at least.

How can it have a 360° view if it’s on the backpack? The only way to have an unobstructed view is to mount a camera on his head lol

The answer to this is really kind of neat. So... yes, if you place a 360 cam on your shoulder and take a picture there will be a blind spot where the wearer's head blocks the camera view. But... we're not using the camera to input a still image. For the most part, the wearer's natural walking gate, the way their body shifts as they walk, allows the camera to get a nearly unobstructed view. Not 100% of the time, but enough that it can keep track of whats going on.

 

[kdonnel]

Well, a child-stranger abduction is very rare, only a very few hundred a year. But 200,000 children are abducted by a non-custodial parent each year, and over 1.5 million restraining orders are issued every year. Point being, while I'm developing this with a specific case-use in mind, its potential usefulness could be much broader.

Again, child abduction for those parents it happens to is a big deal but......

Quotes from: https://www.ncjrs.gov/pdffiles1/ojjdp/196466.pdf

■ An estimated 203,900 children were victims of a family abduction in 1999. Among these, 117,200 were missing from their caretakers, and, of these, an estimated 56,500 were reported to authorities for assistance in locating the children.

■ Forty-three percent of the children who were victims of family abduction were not considered missing by their caretakers because the caretakers knew the children’s whereabouts or were not alarmed by the circumstances (see “Conceptualizing the Problem,” below).

■ It is possible for a child to have been unlawfully removed from custody by a family member, but for that child’s whereabouts to be fully known. Thus, a child can be abducted, but not necessarily missing. This helps put that 200,000 in context.

■ Forty-four percent of family abducted children were younger than age 6. (The device wouldn't work for 44% of kids abducted)

■ Fifty-three percent of family abducted children were abducted by their biological father, and 25 percent were abducted by their biological mother. (78% of non custodial abductions are done by the biological parent who does not have custody. So the device would only be useful in 12% of cases)

■ Forty-six percent of family abducted children were gone less than 1 week, and 21 percent were gone 1 month or more.


That 200,000 child abduction rate is really misleading.

 

[monsterkitty]

I stand by my earlier comment - this is ridiculous.

 

[Mackenzie Click-Mickelson]

I'm not trained is psychology or criminal justice. I am trained in analyzing human movement. In particular I've been using human movement models to improve routing algorithms in self-driving vehicles. The science encompasses determining intent and predicting behavior based on human movement. Things like picking which person in a moving crowd will change direction. So I don't claim any expertise in how a criminal might behave, but I do very well writing code that can determine how most people in a situation are behaving and then pick out the outliers.

Thank you for your information.

Writing computer code is entirely different here and while I think it's really cool that you're involved in self-driving vehicles that type of experience and knowledge does not translate here.

It's irresponsible IMO to go about this without having any training among so many other issues---and I'm not advocating for getting training because I say this nicely and respectfully I highly suggest backing away quickly from this endeavor. It's just wrong on multiple levels.

 

[Squirlz]

I stand by my earlier comment - this is ridiculous.

Look at the posters history. It is like a Burger King menu.

 

[tcufrog]

Impressive tech but it falls under the old adage...

Just because you can do it doesn’t mean you should do it.

 

[PollyannaMom]

My backpack cam (still can't figure a non ominous sounding name for it)

Shoulder Angel?

I really appreciate that you have been coming back and explaining things, and not being defensive. I still don't think you should do it, but I am amazed at the tech.

Several people (including me) have mentioned increasing the parents' fear, but I also think if it became widespread, it would increase everyone's fear. - Can you imagine everyone walking around with their heads down, so as not to be falsely accused of "looking at a child too long"? How many parents are going to go confront someone, and that confrontation goes bad, even though the person wasn't after the child in the first place? How many times, in our current society, have accusations ruined people's careers - before they even got to trial? It's not OK.

 

[Kellykins1218]

I think some people are paranoid and you have way too much time on your hands.

 

[JLTraveling]

I'm not trained is psychology or criminal justice. I am trained in analyzing human movement. In particular I've been using human movement models to improve routing algorithms in self-driving vehicles. The science encompasses determining intent and predicting behavior based on human movement. Things like picking which person in a moving crowd will change direction. So I don't claim any expertise in how a criminal might behave, but I do very well writing code that can determine how most people in a situation are behaving and then pick out the outliers.


Suspicious generally meaning the person was watching the backpack kid more than other people present.

OK, this explains a lot. You have no training in understanding human behavior, but you're good at "picking out the outliers." Sorry, but experience making sure a self-driving car doesn't run over people IN NO WAY translates to determining criminal intent. LOTS of people are drawn to cute kids because they're cute kids, and are often the most interesting thing in the room. Case in point: my cousin graduated from high school a couple of weeks ago. As is true of any high school graduation, most of it was boring. But there was an adorable little boy, maybe 2 years old, sitting in the row in front of us. He kept turning around and smiling and waving, and eventually grabbing for our fingers. Result? Most of my family spent the bulk of the graduation playing with the little boy. His mom was right there, and after the graduation she thanked us for entertaining him. But if he was a "backpack kid," your system certainly would have pegged us as predators for being more interested in the cute kid than the boring speeches. A human (in this case his mother) was able to ascertain that we weren't the least bit dangerous--but would a paranoid mother have been able to deduce that from photos? I doubt it.

Several people (including me) have mentioned increasing the parents' fear, but I also think if it became widespread, it would increase everyone's fear. - Can you imagine everyone walking around with their heads down, so as not to be falsely accused of "looking at a child too long"? How many parents are going to go confront someone, and that confrontation goes bad, even though the person wasn't after the child in the first place? How many times, in our current society, have accusations ruined people's careers - before they even got to trial? It's not OK.

This. My mind immediately goes to every random kid I've ever made silly faces at, and wondering how long it would take till I was accused of wrongdoing. It's extremely disturbing.

 

[MrsPete]

I am trained in analyzing human movement.

Makes me think of something I see every day at school: Our front entrance is made up of four sets of double doors (so eight doors). The kids -- these are high schoolers -- get off the bus and walk like ants towards the middle door -- even if it's cold or rainy. They stop and stand in line to enter that middle door. Lots of teachers do too. A small number of us veer to the left or right and walk straight into the building. I often think, "Why do they do this?"

Again, not really a chore.

I think you don't consider it a chore because you enjoy this type of work and are clearly very "into" this project. I think most people would consider it a chore, one more task to be woven into an already busy lifestyle.

The hardware is very robust. I don't have any interest in getting it rated waterproof, the backpack and its enclosure make it very splash proof.

This matters. About every other week one of my students has a "blow-out" of some sort in a backpack -- maybe a container full of spaghetti, maybe a lid that came off a water bottle.

I presented no false dilemma.

Of course you did. You said this mother /child essentially have two options: 1) Mom remains paranoid and curtails the child's freedom /life, or 2) They use this product. Yes, this is Mom's perception of the truth, but -- in reality -- those aren't the only two options available to her.

Several people (including me) have mentioned increasing the parents' fear, but I also think if it became widespread, it would increase everyone's fear.

That's a fair point. I don't want to be considered "suspicious" because I was reading a little boy's tee-shirt or wondering whether I could copy a little girl's hairstyle on my own daughter.

 

[Colleen27]

I think it is more than a little creepy and absolutely feeding an irrational fear that gets blown out of proportion in our culture. If you could mass produce them, you'd probably make millions.

I think it sounds great! But it would be illegal in my state--you can't record audio or video without consent here. We just use a good old-fashioned cell phone with an app called Life360. I can see exactly where my kids are at any time

Just FYI, two-party consent doesn't generally apply to still photos or video without audio. So this system probably wouldn't violate any privacy/recording laws in any U.S. jurisdiction.

 

[DisMike]

skip

 

[cobright]

Just because you can do it doesn’t mean you should do it.

I tend to agree.

I think some people are paranoid and you have way too much time on your hands.

I'm actually project stacked pretty bad. I've got two power-wheelchairs on my bench and a 15k lb sailboat in my backyard I need back in the water by end of summer. But... some projects get rainy-day status.

Sorry, but experience making sure a self-driving car doesn't run over people IN NO WAY translates to determining criminal intent.

Well, there are some overlaps. In another post I made about the Jetson Nano computer I mentioned how the machine learning and computer vision hardware built into the board were being developed to observe human body language and determine human intent. (Link Here). One application of this is store loss prevention. Most people who have shoplifted something have predictable body language.

LOTS of people are drawn to cute kids because they're cute kids, and are often the most interesting thing in the room.

Reminds me of a Scrobius Pip song, "Thou shalt not think that any male over the age of 30 that plays with a child that is not their own is a peadophile Some people are just nice."

But if he was a "backpack kid," your system certainly would have pegged us as predators for being more interested in the cute kid than the boring speeches.

Actually, if lots of people are keeping an eye on the backpack kid, that behavior normalizes for that particular time and place.

My mind immediately goes to every random kid I've ever made silly faces at, and wondering how long it would take till I was accused of wrongdoing. It's extremely disturbing.

That isn't really how the system works nor is it how I've described it. Now, if you see a stranger's kid and follow that kid down the road a bit, then the next day hang out along that route, and the next day show up at the kid's school, and then over the next several days hang out on the sidewalk near the kids school playground at recess, and on every occasion have your eyes locked on the back of that kid... The system will probably generate a report with your pic so the parents can tell it if you are a relative or family friend.

but -- in reality -- those aren't the only two options available to her.

But I wasn't pondering which choice the mother should make.

even if it's cold or rainy. They stop and stand in line to enter that middle door. Lots of teachers do too. A small number of us veer to the left or right and walk straight into the building. I often think, "Why do they do this?"

Can see this at WDW entrances as well.

I think you don't consider it a chore because you enjoy this type of work and are clearly very "into" this project. I think most people would consider it a chore, one more task to be woven into an already busy lifestyle.

I said it wasn't much of a chore because that phase of the system setup only lasts about two weeks and even then it is only a matter of looking at a page or two with a dozen photos on each. It literally takes 3 minutes each night. After setup is done, a report review happens every couple of days and involves looking at a couple of photos.

That's a fair point. I don't want to be considered "suspicious" because I was reading a little boy's tee-shirt or wondering whether I could copy a little girl's hairstyle on my own daughter.

The system currently doesn't decide a person is suspicious based on a single interaction with that person. As I learn more about coding for this platform and more libraries become available there are more complicated behaviors that a person might exhibit that the system could detect and use to make a determination sooner, but the idea here is that it takes a long view on the surroundings and doesn't reach any decisions based on a single action.

So as I ponder what I've managed with this so far and how the technology might be used in ways maybe less controversial... How about this... Just spitballing...

A device that...

• lets you know whenever someone has taken your picture?
• lets you know if someone is trying to get your attention? In addition to general utility, some people with ASD are unable to read body language and non-verbal cues.
• lets you know if someone is talking about you to someone else? (less reliable but based on hand gestures, direction of the person's vision... this could be done)
• lets you know if people or even specific people around you are talking to you? Some people with sensory issues would prefer to wear a noise reduction device but fear missing communication directed at them.
• lets you know the name of the people around you if you know them. Some people are simply horrible at remembering names. I'm great at remembering names but I have a kind of face blindness and if people I've known my whole life all show up wearing similar clothes (say at a wedding or funeral) I will simply not reliably tell one from the other.

 

[cobright]

Look at the posters history. It is like a Burger King menu.

Interesting observation. It doesn't seem like a complimentary simile at all and yet it lacks the conviction to actually criticize anything relating to the topic of this thread.

But yes, by all means rummage my history... You'll find I

• talk mostly about mobility related technology, wheelchairs and ECVs. My primary occupation is developing better mobility devices and better controls for them.
• give a lot advice about coping with anxiety and panic disorders, from the perspective of someone who went from being crippled by them to someone who is back in the world.
• answer a few questions, always with practical advice.
• chime in occasionally with fairly moderate centrist political views on subjects other people bring up.
• My love for Disney souvenir mugs (tiki especially)

That takes me a little over a year back ... 10 pages of history ... so you click 'view older results' and ...

• Pretty much same thing
• a question then update on finding a Ramos Gin Fizz in WDW
• I give negative(ish) Carthay Circle review
• I comment in a thread about open relationships that my wife and I have a platonic relationship with a woman I have known for decades and we often travel together. I suppose that might be the Burger King-iest thing I have mentioned.

Or maybe just formulate a real argument? I'm happy to receive actual criticism. If the capabilities of the computer system seem dubious to you I'll link some demos of the technology I'm using from another post I made ... here ... the yolov demos are really pretty slick examples of object identification in realtime. As I hope I've mentioned by now, 90% of the coding for this project comes from existing libraries. I'm not claiming to have invented some super computer, just mixing yolov and tensorflow up in some new ways.

Or whatever. Potshots from the peanut is a lot less effort I suppose.