Lost item finding Ai camera

[Levin Witt]

Hello everyone,
I'm trying to validate an idea I've been thinking about for some time. I wanted to know what you ever consider buying or see someone buying a camera that you could put somewhere, (usually in your house), that uses an object detection ai to catalog all the items it sees, in order for you to be able to search for a specific item should you loose it. You could do this over your phone via an app maybe, where you search for sunglasses say, and any instances of sunglasses are returned to you to find where you put them. If you put them in something like a box or a bag, it would then have to be able to track that, essentially keeping track of where everything it sees even just once at all times. You might need multiple cameras to get all angles. Security is a big issue, since everything in your house would be filmed, which maybe could be solved with encryption. I think I could prototype it with something like a raspberry pi. Also I would hire some ai freelancer to help me with the coding since I'm not nearly good enough at it to pull this of (even though I know some things) by just myself (if it's even technically possible-this remains to be seen).

Thank you, I look foreword to your feedback!

 

Dislike ads? Remove them and support the forum: Subscribe to Fastlane Insiders.

[JeremyCreed99]

Hello everyone,
I'm trying to validate an idea I've been thinking about for some time. I wanted to know what you ever consider buying or see someone buying a camera that you could put somewhere, (usually in your house), that uses an object detection ai to catalog all the items it sees, in order for you to be able to search for a specific item should you loose it. You could do this over your phone via an app maybe, where you search for sunglasses say, and any instances of sunglasses are returned to you to find where you put them. If you put them in something like a box or a bag, it would then have to be able to track that, essentially keeping track of where everything it sees even just once at all times. You might need multiple cameras to get all angles. Security is a big issue, since everything in your house would be filmed, which maybe could be solved with encryption. I think I could prototype it with something like a raspberry pi. Also I would hire some ai freelancer to help me with the coding since I'm not nearly good enough at it to pull this of (even though I know some things) by just myself (if it's even technically possible-this remains to be seen).

Thank you, I look foreword to your feedback!

Sounds like a good app!

 

[Bekit]

I can not tell you how many times I've wished that I could press CTRL + F in the real world to find an item.

 

[Martin.G]

It's a good idea, but I see at least a problem. Do you have to have a camera in the room all the time? How many people are willing to have one and pay for that?. Maybe I can use an app that do something similar but in real time with it own camera and all the process in the same hardware.

Also, it could work if for example is a small device that I don't mind having it in the room, but at the end of the day you need a lot of resolution and power to process the information. In general, you need a GPU to do some interesting things in machine learning. For example Tesla with its expensive hardware to make the necessary calculation to not crash.

I think your idea is to process in a server, but in that case you have two problems. First, you need a lot of hardware to process the information too, and second you need a client who is willing to send that information. I don't know, I see a lot of problems. It doesn't matter the encryption. People are complaining about Facebook having and using their information, like what they eat or when they go to the bathroom, and you want to record a room.

What I would do is try to validate that hypothesis, and other that you have. For that I recommend you read a couple of books about validation (Sprint: How to Solve Big Problems and Test New Ideas in Just Five Days, for example). The idea is to create a prototype and test with different people. You don't need to code, maybe a camera and a couple of people in other room watching and doing the ML part.

The idea is nos enough to know if people like it. You are going to have someone tell you that is a great idea but at the end of the day there are not going to pay you probably. You need to validate fast and cheap, and iterate from that to change it the thing you saw. That way you are going to create the solution and not what you think it is.

 

Dislike ads? Remove them and support the forum: Subscribe to Fastlane Insiders.

[loop101]

Hello everyone,
I'm trying to validate an idea I've been thinking about for some time. I wanted to know what you ever consider buying or see someone buying a camera that you could put somewhere, (usually in your house), that uses an object detection ai to catalog all the items it sees, in order for you to be able to search for a specific item should you loose it. You could do this over your phone via an app maybe, where you search for sunglasses say, and any instances of sunglasses are returned to you to find where you put them. If you put them in something like a box or a bag, it would then have to be able to track that, essentially keeping track of where everything it sees even just once at all times. You might need multiple cameras to get all angles. Security is a big issue, since everything in your house would be filmed, which maybe could be solved with encryption. I think I could prototype it with something like a raspberry pi. Also I would hire some ai freelancer to help me with the coding since I'm not nearly good enough at it to pull this of (even though I know some things) by just myself (if it's even technically possible-this remains to be seen).

Thank you, I look foreword to your feedback!

You could just track where the user goes in the house, and tell them to search those locations for whatever they lost.

 

[Raja]

I suggest you use 360 home camera with wifi enabled, i Don't know much about this step send data to gcp, process it on cloud, interface via app.

Though you need to optimize it, like run when required

 

[Stargazer]

Seems logical but would you not need a camera in every room and hallway on 24/7?

Which would be costly plus do you want a camera 24/7 in every room?

Like most things the corporate space is where you would start as they can afford it.. And then costs would come down.

Airports lose luggage. This could minimise their compensation payouts.

Warehouses routinely do Stock Takes. And usually there is a discrepancy which if within a parameter is written off but even that can add up if it is a massive organisation.

I would start there if I was you.

Dan

 

Dislike ads? Remove them and support the forum: Subscribe to Fastlane Insiders.

[Chibbs]

I have some experience with these AI products using cameras at my day job. I would suggest finding a specific application or object with which to track. A general tracker is going to take a lot of training and your accuracy will suffer

 

[Raja]

Seems logical but would you not need a camera in every room and hallway on 24/7?

Which would be costly plus do you want a camera 24/7 in every room?

Like most things the corporate space is where you would start as they can afford it.. And then costs would come down.

Airports lose luggage. This could minimise their compensation payouts.

Warehouses routinely do Stock Takes. And usually there is a discrepancy which if within a parameter is written off but even that can add up if it is a massive organisation.

I would start there if I was you.

Dan

Hey Dan,
360 home security camera are cheap af, may be not af but cheap. You can get it from china or use mi product in or under 30 usd.

 

[Levin Witt]

Seems logical but would you not need a camera in every room and hallway on 24/7?

Which would be costly plus do you want a camera 24/7 in every room?

Like most things the corporate space is where you would start as they can afford it.. And then costs would come down.

Airports lose luggage. This could minimise their compensation payouts.

Warehouses routinely do Stock Takes. And usually there is a discrepancy which if within a parameter is written off but even that can add up if it is a massive organisation.

I would start there if I was you.

Dan

I would say this is good, but it is already being done in places like airports or factories. I could try to find a spot there but my thought was doing it for private consumers which is not very wide spread at all. But thank you very much for your input, I will do some research.

 

Dislike ads? Remove them and support the forum: Subscribe to Fastlane Insiders.

[Levin Witt]

I have some experience with these AI products using cameras at my day job. I would suggest finding a specific application or object with which to track. A general tracker is going to take a lot of training and your accuracy will suffer

Sounds like a good idea, i'll look into it!

 

[Levin Witt]

It's a good idea, but I see at least a problem. Do you have to have a camera in the room all the time? How many people are willing to have one and pay for that?. Maybe I can use an app that do something similar but in real time with it own camera and all the process in the same hardware.

Also, it could work if for example is a small device that I don't mind having it in the room, but at the end of the day you need a lot of resolution and power to process the information. In general, you need a GPU to do some interesting things in machine learning. For example Tesla with its expensive hardware to make the necessary calculation to not crash.

I think your idea is to process in a server, but in that case you have two problems. First, you need a lot of hardware to process the information too, and second you need a client who is willing to send that information. I don't know, I see a lot of problems. It doesn't matter the encryption. People are complaining about Facebook having and using their information, like what they eat or when they go to the bathroom, and you want to record a room.

What I would do is try to validate that hypothesis, and other that you have. For that I recommend you read a couple of books about validation (Sprint: How to Solve Big Problems and Test New Ideas in Just Five Days, for example). The idea is to create a prototype and test with different people. You don't need to code, maybe a camera and a couple of people in other room watching and doing the ML part.

The idea is nos enough to know if people like it. You are going to have someone tell you that is a great idea but at the end of the day there are not going to pay you probably. You need to validate fast and cheap, and iterate from that to change it the thing you saw. That way you are going to create the solution and not what you think it is.

Yes I know people would never accept having their information sent out, I wouldn't myself (that is if facebook isn't already hacking my laptop camera and watching me rn). The idea was to do all the processing and ai stuff locally, ideally in the camera itself. That way you wouldn't have a feed going anywhere, except when you interact with the app on your phone. There would be no central server. Tensor flow lite can be used to run ai on edge devices like a raspberry pi. But it would probably require a lot more sophisticated hardware than a Pi. Idk what the best way to solve this is, as i'm not an engineer .

 

[Equilibrium]

I'm your target audience. But I wouldn't buy this product.
Let me explain, and offer my $00.2 to expand on this.
Every...single.. The item I own (excluding clothes) has a label glued onto it.

------
link to a Google form
Item Number
Pin (required to submit form)
QR Code Linking the Item Number
------

In a Google Spreadsheet is a list of every item I own along with
Prices of items, Actions on items.
As an example, next month at 6:45 a script will remind me to replace the fishtank filter.

Each month I check a list of items, then I walk around with a scanner.
beep. Item 016 beep 009..

Why do I do this?
For convenience, it provides of linking a real-world item with any tasks associated with it. I can give a binder to a house sitter with in-depth details. I can check the value of lost items from luggage. I can decide if something should be given away even if that item is sitting in a box in the attic.

it was absolutely bloody hell setting the system up, and I intend to never do it again in my life.
(adding new tags is painless.)

Your idea requires the same process. First, finding someone like myself who frankly gives a cow about finding lost eyeglasses and then convening them to use tags on objects.
THEN getting a camera that can read those tags, without violating the privacy, AND dealing with all the thousands of complaints about network connectivity. Unless you want to use just the objects, which requires a lot of training for the models.

Again I'm your target audience. But had I known ahead of time the level of life upheaval required to track said items? Hahaha I'd have never even started it. Additionally, I'm a minimalist, meaning I have few items, and this was STILL hard. Finally, speaking from experience, you really..really don't want to deal with network issues remotely.

Now should you lay around with it, see if you can figure out the AI? sure! get the experience. Maybe you will come up with an amazing idea that utilized that first demo.

After all that's what Snapchat did with Qr Codes, snapping up a startup that was trying to affix them to real-life objects.

PS
Airports lose luggage. This could minimize their compensation payouts.
I can attest that having a traceable list of all items in my luggage case 100% improved the speed of receiving a payout. It also ensured the return of a lost cell phone in a different state.

 

Dislike ads? Remove them and support the forum: Subscribe to Fastlane Insiders.

[Martin.G]

Yes I know people would never accept having their information sent out, I wouldn't myself (that is if facebook isn't already hacking my laptop camera and watching me rn). The idea was to do all the processing and ai stuff locally, ideally in the camera itself. That way you wouldn't have a feed going anywhere, except when you interact with the app on your phone. There would be no central server. Tensor flow lite can be used to run ai on edge devices like a raspberry pi. But it would probably require a lot more sophisticated hardware than a Pi. Idk what the best way to solve this is, as i'm not an engineer .

What you want to do it remind me what Amazon do in his physical stores. You are going to hire more than a freelance programmer to create something like that. That's why I advise you before, because it's not going to be cheap and it's better you be sure that it's the right solution.