I want to do a little bit of secret recording. My ideal is I can have a program that parses the video feed and recognises what I view. Let's say I want to program it for when I go food shopping. I want it to differentiate an apple from an orange, and recognise that one apple is different from two apples. It also needs to be able to recognise the apple from different angles.
Am I deluded to think this is possible without some google style neural network? Remember, I want to do this with a spy camera.
Watched this video. Will explore his channel. Thanks.
I am not a computer expert, forgive me for stupid questions or terminology: Is this easier for his algorithm because it is only has to recognise computer visuals rather than a camera feed?
Something that does simply the wanted camera feed is little environmental change. I will likely be sitting in one position and only move upper body/head.
Brayden Foster
I am concerned that if I do wireless sending of the information I will have the communications intercepted. I think having wires from the camera to an attached computer is possible. Can information 'leak' from wires and be picked up by those that wish to listen? I'll be in an environment where plenty of people are using phones, digital watches and there are multiple computers in the room. Will my signals along the wire be clearly noticeable to people that would wish to listen in?
Caleb Campbell
If you can pull this off and publish your methods you will have earned yourself a PhD in machine vision. The problem you're describing is that hard.
David Reed
I am sure this is bait, but I really want to know where you are planning to go with this? What kind of sensitive data would you obtain by literally comparing apples to oranges?
Ethan Kelly
Nah, it's not James Bond save the world shit. A simple money making scheme. If I can get some information, I get money. If I get noticed, I will be deported and the money making opportunity is gone for me. I don't want to be noticed.
Not apples and oranges. Standardised documents. Could I get the machine to recognise the standardised documents by taking videos of them before hand so it can recognise them in the wild? I will be clearly seeing the documents, from 1-2m away. Not like it has to be seen past a sea of people from 10m etc
>I will be clearly seeing the documents, from 1-2m away. Not like it has to be seen past a sea of people from 10m etc why do you need AI for this. just record it with the pen
Angel Carter
Ahaha good luck
John Harris
Because I want it to process it in real time. That's where the money is.
The pen is only for video feed. Processing will occur on a computer attached to my body. I don't have a clue how powerful the computer needs be.
William Diaz
>Processing will occur on a computer attached to my body. I don't have a clue how powerful the computer needs be. I have doubts that ARM based phone will be good enough but it's certainly worth a try. The one catch that you shouldn't forget is to use a neural network you need a gigantic training set, tens of thousands of labeled pictures.
Matthew King
Labelling pictures should be fine. I'm flying and in hotels frequently. Something to do while airport, plane, on the shitter etc
tens of thousands seems far more than I'd heard. Heard about 500-1000 for each object. Only got about 15 objects to do. Don't want it to identify anything else.
Austin Parker
I like my sister too.
Julian Walker
Why would I need object recognition for a sister? Not like I need a computer to show me what a titties
David Peterson
>I want it to differentiate an apple from an orange, and recognise that one apple is different from two apples. Sure thing, I want to go to the mars and become immortal. Any other crazy dreams?
We are a few centuries away from computers being able distinguish and group arbitrary objects.
>recognise the standardised documents With a few million dollars and a small development team you could probably create such software in a few years time. Or you could train a neural network, good luck with that, you could probably achieve acceptable results quite fast (a couple of months?) if you know what you are doing and have an acceptable training set. Keep in mind that if you want to recognize very small differences, such as the layout of documents (I have no clue what you mean by "standardized document") you will need a very large sample size. Processing is no problem a phone should be enough.
Cooper Gonzalez
>My ideal is I can have a program that parses the video feed and recognises what I view. kek >and recognise that one apple is different from two apples. top fuckin kek you're clueless.
Jonathan Parker
>show me what a titties If this were still old Sup Forums this would have been a meme.
Brody Jones
>Keep in mind that if you want to recognize very small differences, such as the layout of documents (I have no clue what you mean by "standardized document") you will need a very large sample size. Thanks. Cause yes, it is a few lines of difference between each. To us it looks completely different, but maybe not to a computer.
Yes, yes I am clueless. Hence I came here to find the basics rather than be out there doing the business I want to do.
