Blind and Visually Impaired Community
↑
11
↓
I made this tool that lets your Android phone tell you what the camera sees as you wave it around. It speaks and vibrates when it is confident. (play.google.com)
submitted by ieee8023
Lion_the_Bunny 4 points 7y ago
This is fucking cool. That is all.
fastfinge 2 points 7y ago
Any plans for an IOS app? I, and I think the vast majority of totally blind folks, have IOS instead of Android, because the accessibility of IOS is leaps and bounds ahead of Android.
battez 2 points 7y ago
AiPoly is a commercial one that has been released I think
fastfinge 2 points 7y ago
Yeah, I've played with it, and it seems pretty similar.
ieee8023 [OP] 2 points 7y ago
I only know Android development. Some engineers have said they want to help and maybe they know iOS.
Marconius 1 points 7y ago
I second a port of this to iOS. Voiceover is vastly superior to talk back in every way shape and form, and it will get much more marketshare and spread around the blind community that way. It would be great if it had a live OCR application in future updates, although I know the processing for doing that live is extremely intense.
geoffisblind 1 points 7y ago
Seems interesting, I just downloaded it, I'll play around with it in a bit when I get some time. Thanks for sharing!
Edit: This is really cool, I'm definitely going to keep it around, do you have plans to expand it any further? I noticed it said it is in Beta.
Edit: This is really cool, I'm definitely going to keep it around, do you have plans to expand it any further? I noticed it said it is in Beta.
ieee8023 [OP] 1 points 7y ago
It's Beta because I made it 4 days ago and there are most likely bugs.
I hope to do research to make the network more accurate, faster, and have more labels. It could potentially know 27,000 labels but the state of the art accuracy is ~30%. For the 1000 labels it currently knows it could do better but it would take longer between readings ~5-10s instead of 1s currently.
I hope to do research to make the network more accurate, faster, and have more labels. It could potentially know 27,000 labels but the state of the art accuracy is ~30%. For the 1000 labels it currently knows it could do better but it would take longer between readings ~5-10s instead of 1s currently.
geoffisblind 2 points 7y ago
Well I've been playing around with it and it works really well so far, it definitely recognizes more things than I was expecting, how are you doing that? I find when I tried to use it on cloth things it kind of got thrown for a loop, but everything else has been incredibly accurate against the things I was testing it on. I love that the Android users are getting some love, we are usually left in the cold when it comes to accessibility apps so it's nice to have something. Great job :)
ieee8023 [OP] 2 points 7y ago
Great! I'm glad it is working well for you!
It uses a Convolutional Neural Network implemented with mxnet.
It uses a Convolutional Neural Network implemented with mxnet.
This nonprofit website is run by volunteers.
Please contribute if you can. Thank you!
Our mission is to provide everyone with access to large-
scale community websites for the good of humanity.
Without ads, without tracking, without greed.
scale community websites for the good of humanity.
Without ads, without tracking, without greed.
©2023 HumbleCat Inc • HumbleCat is a 501(c)3 nonprofit based in Michigan, USA.