PARSEE:World's First FREE Glasses for Blind People

[daniel19210]

Hi everyone.
A project I've been working on for past year just hit indiegogo. If you have any questions I'll be happy to answer.
Please spread the word further. Thanks!

https://www.indiegogo.com/projects/parsee-world-s-first-free-glasses-for-blind-people--3/x/13909914#/

[Xenoamor]

I like the idea but I have a few questions:

1) What is the money you raise going to be spent on. You mention consultation with engineers?
2) As a non-profit organisation are you going to make the project open source
3) Why are you applying to patent this?

[edy]

Hi,

Yes this is a fantastic idea. I was wondering if you managed to create even a rudimentary prototype (say using a smartphone camera/app or laptop with webcam)? It seems to me the device will have to be cloud-connected to tap into enough computing power to handle the recognition (much like Google Goggles). Will the user be required to be connected to a smartphone, or will the glasses have everything built in?

While I applaud your efforts, I think the relatively low amount of money being raised for the challenges ahead, as well as the relatively cheap cost you are offering for these glasses, makes me concerned. Since the prototype hardware itself should be very inexpensive, I see all of the work involved being in software. It should be possible to at least build a proof of concept for next to nothing. [edit:] But even then... the costs in miniaturizing hardware can add up, and finding a manufacturer to make even a small run would be prohibitively expensive.

For example, you can strap a smartphone to your head with the camera on, and have it recognize objects with cloud-based recognition (you could have it link up and send a video feed to a server in your network which does all the grunt-work and then relays results back as text , which the app will convert using text-to-speech API's). [edit:] I see in the video you are using an IP-based camera... Is it broadcasting the image to a PC-based software that is parsing the video for recognizable objects?

Anyways, it sounds a lot like a video real-time version of Google Goggles. You should talk to the people at Google about your idea, I'm sure you can leverage their technology and maybe they have an API to save you a lot of work in creating a recognition engine yourself. Or if you have, perhaps Google can use it.

I look forward to seeing what you develop. Good luck with this ambitious but much-needed project. 

[daniel19210]

I like the idea but I have a few questions:

1) What is the money you raise going to be spent on. You mention consultation with engineers?
2) As a non-profit organisation are you going to make the project open source
3) Why are you applying to patent this?

1) We need to conduct more research and customize product according to blind and visually impaired people needs. Research and development unfortunately costs a lot. Finally according to our vision we want to distribute our eyewear to visually disabled for free. Consequently we expect significant demand and are sure we won’t be able to cover material and shipping costs – therefore we need your support.
2) We plan that some parts will be open sourced but the most advanced algorithms will be protected.
3) With reference to previous question - only parts, not whole solution. Some algorithms are masterpieces of our engineers and we would like to protect them.

[daniel19210]

Hi,

Yes this is a fantastic idea. I was wondering if you managed to create even a rudimentary prototype (say using a smartphone camera/app or laptop with webcam)? It seems to me the device will have to be cloud-connected to tap into enough computing power to handle the recognition (much like Google Goggles). Will the user be required to be connected to a smartphone, or will the glasses have everything built in?

While I applaud your efforts, I think the relatively low amount of money being raised for the challenges ahead, as well as the relatively cheap cost you are offering for these glasses, makes me concerned. Since the prototype hardware itself should be very inexpensive, I see all of the work involved being in software. It should be possible to at least build a proof of concept for next to nothing. [edit:] But even then... the costs in miniaturizing hardware can add up, and finding a manufacturer to make even a small run would be prohibitively expensive.

For example, you can strap a smartphone to your head with the camera on, and have it recognize objects with cloud-based recognition (you could have it link up and send a video feed to a server in your network which does all the grunt-work and then relays results back as text , which the app will convert using text-to-speech API's). [edit:] I see in the video you are using an IP-based camera... Is it broadcasting the image to a PC-based software that is parsing the video for recognizable objects?

Anyways, it sounds a lot like a video real-time version of Google Goggles. You should talk to the people at Google about your idea, I'm sure you can leverage their technology and maybe they have an API to save you a lot of work in creating a recognition engine yourself. Or if you have, perhaps Google can use it.

I look forward to seeing what you develop. Good luck with this ambitious but much-needed project. 

We have a working prototype that you can see in pictures and videos. The one with bigger frames. All the processing is done by our application installed on an Android smartphone. We are planning to leave it that way because modern phones have lots of processing power that is not fully utilized by it's users. Offloading all the image recognition to a portable device that almost everyone owns by now, allows us to use cheaper components and extend battery life.

[daniel19210]

I'm not sure what gives you the balls to say you have advanced enough algorithm to recognize everything from a crappy IP camera.
First, you need at least 60fps, preferably 100fps refresh rate plus global shutter sensor to clearly capture moving objects.
Second, if all your valuable algorithm is in an app, then how do you manage to dump 60fps+ HD video through WiFi to your phone?
Third, when the phone itself is used in WiFi mode, you can not create AdHoc network to the camera, you need the round trip bandwidth (camera-router-phone) to be sufficient, which varies a lot, so is not reliable.
Fourth, there is no true intelligence property protection in Android, decompiling Java byte code is way easier than decompiling C/C++ code. If you chose to use C/C++, then how can you effectively support both platforms? Cortex A6/A8/A9/A15/A53/A57/Atom, etc. Each processor family has a slightly different acceleration ISA to make code more efficient, which you might want to use.
Fifth, the main problem, how do you plan to recognize everything? If you are talking about OpenCV+tesseract, then save your time, it is far from the same level of human eyes. If you want to conduct your own research on CV algorithm, then the money is far from enough. Remember you are competing with major universities' labs and big companies.

Right now we're sending 720p at 30fps. Sending even higher resolutions and frame rates will not be a problem thanks to encoding.
We don't need to run connections through a router. Glasses are connected directly to the phone.
It is true that android apps can be easily decompiled and that's why we want to patent our algorithms. I'm not sure if you're familiar with how patents work but in order to apply for one you need to disclose your algorithm and after your patent is applied the algorithm becomes public. No one can use it though because it's illegal to use patented technology for commercial purposes without patent holder's permission. Considering that our algorithm is going to be publicly described there's no point to take any measures to prevent people from extracting it from our app. We don't need to worry about protecting it either because that will be taken care of by the patent.
I cannot answer to your last concern as that would mean disclosing classified information. I can assure you though that we are aware of that problem. So far even an average smartphone has enough processing power to run our algorithms.

[daniel19210]

I'm not sure if you're familiar with how patents work but in order to apply for one you need to disclose your algorithm and after your patent is applied the algorithm becomes public. No one can use it though because it's illegal to use patented technology for commercial purposes without patent holder's permission.

Software patent is only recognized in a handful of countries, such as the USA and Korea, most of the world does not recognize software patent. The EU and most Asian governments explicitly excluded software from being patentable.

Even your patent is valid, it is very easy to bypass it, after all image processing and recognizing is a very active field of research, and new ideas pop up every single day. If someone wants to bypass your patent, hire a few PhDs and work for a week, they will have a new method that breaks your patent chain. A patent is violated only if successive points in your patent are violated in order. If someone breaks the chain at its weakest point, the patent is meaningless.

We are satisfied with the number of countries acknowledging software patents. It's a bit more than handful.

And yes. Patents are unfortunately a very vague thing. We'll prepare ours with the help of professionals to reduce the risk of our patent being bypassed. Of course that doesn't mean it will be 100% secure. Again we're satisfied with the level of protection we'll get from patenting our work.

[edavid]

Not the best product name, unless you happen to be Zoroastrians: https://en.wikipedia.org/wiki/Parsi

[daniel19210]

Not the best product name, unless you happen to be Zoroastrians: https://en.wikipedia.org/wiki/Parsi

Thanks for letting us know. We won't be changing our name though.

[ez24]

Not the best product name, unless you happen to be Zoroastrians: https://en.wikipedia.org/wiki/Parsi

Thanks for letting us know. We won't be changing our name though.

I have seen other non-English speaking countries picking an English name that is not good and they always refused to change it and crashed.

My suggestion is to spell it this way :  ParSee

the word See is easier to get and has less connection to history

[rrinker]

 Though it is not true that the Chevy Nova failed to sell in Mexico because No Va in Spanish is Not Going.

[Delta]

Though it is not true that the Chevy Nova failed to sell in Mexico because No Va in Spanish is Not Going.

Can any French forum members confirm or deny that the Toyota MR2 was re-badged in France, because "MR2" is pronounced as "Shit 2"...?

[rrinker]

 There seem to be as many sources claiming that they changed the name to MR in France as there are debunking the Nova one, so this is quite plausible.

[edavid]

My suggestion is to spell it this way :  ParSee

Umm, that is how they spell it.  It's an archaic spelling in English, but also what English speakers would recognize, due to Kipling.  It's not in any way a bad word (unless you hate Kipling), but kind of distracting, since it has nothing to do with the product.

[ez24]

My suggestion is to spell it this way :  ParSee

Umm, that is how they spell it.  It's an archaic spelling in English, but also what English speakers would recognize, due to Kipling.  It's not in any way a bad word (unless you hate Kipling), but kind of distracting, since it has nothing to do with the product.

From their "site"  I see  "Parsee"  and "PARSEE"      I suggest  "ParSee"   and See is related to Blind

Their product is related to blindness so if you can See you are not blind.  ie related

Companies do not change their names based on members suggestions.  My suggest is to capitalized the "S".  No name change.

I can just guess that "Par"  means "part"  like in partial sight ?

FYI there was a story on NBC tonight about a device that helps blind people but costs $15,000  (I am sure it has more capability)

 

[timb]

My suggestion is to spell it this way :  ParSee

Umm, that is how they spell it.  It's an archaic spelling in English, but also what English speakers would recognize, due to Kipling.  It's not in any way a bad word (unless you hate Kipling), but kind of distracting, since it has nothing to do with the product.

From their "site"  I see  "Parsee"  and "PARSEE"      I suggest  "ParSee"   and See is related to Blind

Their product is related to blindness so if you can See you are not blind.  ie related

Companies do not change their names based on members suggestions.  My suggest is to capitalized the "S".  No name change.

I can just guess that "Par"  means "part"  like in partial sight ?

FYI there was a story on NBC tonight about a device that helps blind people but costs $15,000  (I am sure it has more capability)

 

What you're describing actually has a name! It's called medial capitals, or more frequently: InterCaps (if you're an English Major) and CamelCase (if you've got a BSE). {If you've got a BSE with a minor in English, you call them InterCamels!}