Google Vision APIs for Android provide face detection and barcode scanning | Hacker News Reader

Google Vision APIs for Android provide face detection and barcode scanning

developers.google.com

Would be nice if they provided more information about their implementation. I currently pay for a barcode API that doesn't have the dumb binary off/on pixel logic of the free version of zxing almost everyone else uses. It's just better UX for users if their bar codes can be recognized even with out of focus images, for example. There's nothing on this site about if the Google library is better or worse than popular open source offerings, however.

which api are you using? im currently looking into the same topic...

Last I looked, Clef, who rely on barcode scanning as the basis for their entire product, were using ZBar[0] on Android. It's a C library. I think it's unmaintained these days though. No commits on sourceforge for 3 years.

[0] http://zbar.sourceforge.net/index.html

I wish this could be used in a broader context without involving Android.

What is the performance of the barcode scanning? The commercial Scandit crossplatform lib is amazing (see videos) www.scandit.com

Wait... what? Why is this part of the Google APIs framework (closed source) and not part of the Android Open Source Project?

Here we go, folks.

Because this is higher level in terms of functionality than OS. Android API lets you access hardware, e.g., get a photo from camera, establish a network connection.

Processing data from hardware is higher level.

I didn't say make it part of the OS, I said part of the Android Open Source Project. The platform SDK (applications, services, UI, media playback, notifications etc. etc.) and the support library are both extremely high level and part of AOSP.

Yeah, it should really be open source. For face detection, it's really a closed source replacement for what's been in Android from the start: http://developer.android.com/reference/android/media/FaceDet...

There are many applications of this stuff outside of Android though. I agree that it's a little weird for it not to be incorporated in AOSP, but it could be used on Chrome OS. Heck, maybe even iOS? I think Google is just trying to keep the door open for developers on different platforms.

Because that way Google gets to decide what uses you put its AI to. Recognizing faces and barcodes? Sure. Want to ask for the AI's help in recognizing business opportunities in click data? Sorry, that functionality is reserved for Google.

Because ZXing Barcode Scanner works anywhere, so Google had to fix it so barcodes can only be scanned on proprietary Google phones.

because the developers would be stuck with the version the manufacturer ships with the device, this way all users are on the same version. they could still open source it though

Probably because they didn't want to open source it?

There's still a lot of secret sauce in this general domain.

Good direction.

I wonder how it compares to using OpenCV other on top of much nicer API and doing everything yourself.

i think this only works on snapdragon processors with fastcv. i tested this on snapdragon dev kits 2 years ago.

any information about licensing the API ? cost-wise ? I couldn't find it in the site

I'm pretty sure processing happens on the device, it's not a web API. No cost.

Totally free, you can use it on any device that has Google Play Services.

Reply on news.ycombinator.com