I'm looking at developing an app that could benefit from having a image recognition system. I've seen this sort of thing in iPhone and Android apps. Take a picture of a book and the app takes you to Amazon where you can find that book. I'm not looking for general image recognition, but more the ability to pick a single image out of a library of about 10k images.
Any ideas of what services are available for this sort of thing?
Google Goggles does something similar to Amazon Remembers. It uses OCR if text can be identified and they want to use it with the similar image search from Google Images. I think they generate some kind of hash for an image with the feature that if the images are similar the images are similar to.
My best guess would be try to start with the character recognition and do a text search for the title of your card. This means your user has to make a very clear image maybe even in a specific position. But for a first application this would be great already. As somebody playing magic I would buy the tool for trading and cataloging my cards.
Actually, while short of getting an actual Amazon employee to tell you there is no way to confirm this, I am fairly certain that the Amazon Remembers feature you refer to is actually the work of crowd sourcing- using lots of people combing through data to make it appear like a computer is doing it. I think they may actually be using there own Mechanical Turk system.
Edit: Also, I found this SO question that might interest you. It is specifically for playing cards, but some of the answers (such as the machine learning example) can be modified to be more helpful for what you want to do with magic cards.
Related
Think you ideally needed to mark up or censor a list of keywords in visual output systemwide, yet cant require to root devices.
That still works on websites through browser plugins.
But is it any thinkable to mess with popular apps like whatsapp, facebook, (one at a time) ?
Reading: I know it is possible to read/change some text inputs yet not generally/all? http://developer.android.com/training/accessibility/service.html
A universal way for markup could be determining screen coordinate positions of contents by OCR and set transparent overlays on the fly + algined smooth with scrolling, just not convinced how well this can both work and be battery efficient (we could cope with low accuracy in text recognition)
I'm adding all my reputation as a bounty.
Laying out a good way for any one popular app (top 20 social apps) qualifies as an accepted answer!
Laying out solution for "1." only but for two or more apps also qualifies.
Showing specifically why/where it will work with one app but not with another also qualifies as an accepted answer.
You cannot modify the visual output of an app. The closest thing you can do to accomplish this is like what Facebook Messenger and LastPass use, which is a feature that allows them to draw over the top of other apps, and LastPass specifically can also read the contents of another app via accessibility permissions. However, that just allows them to draw over the existing apps, and in the case of LastPass, fill out text into input fields. Even with the above options, I don't think you'll be able to accomplish what you're looking to do.
I don't think it's possible. Each app on the Android runs in its own sandboxed environment. You can only communicate through Intents with other applications. So unless they're listening for bad Intents(which I hope they're not), you can't really do anything to those applications.
I'm looking at the possibility of creating an app which will, through the camera on an iphone / ipad / etc., recognise a name on a headstone in a cemetery. It would then match the image (or recognised text) against a stored image / text in a database and call the specific record from that database.
So for example, there is a database of soldiers who are buried in a particular cemetery. All headstones are laid out in the same format (my avatar is one). If I go to that cemetery with the database installed on (eg) an iPad, and point the camera at a headstone, the soldier's information will be retrieved and displayed on the screen. I imagine something like the system for redeeming an iTunes number, but am not sure. This would 'personalise' the headstone by providing much more detail on the soldier's life and ultimate fate, and could be very useful to people visiting cemeteries for the 100th anniversary of WW1 commemorations. Any suggestions gratefully accepted. Thanks.
Cool idea, but I'm not sure why you have come to stack overflow for it as it is an idea, but not a question that people can give you a definitive answer, perhaps this would be better suited on a forum somewhere?
I would suggest you research optical character recognition and using OpenCV to process images as you might be able to achieve the results you want. But aside from that I can't really give you much more information.
Good luck with the project, sounds interesting.
Google provides a variety of 'cards' for Google Now (http://www.google.com/landing/now/). Is it possible to create your own cards? The system looks pretty modular, but I haven't found any documentation or instructions to do so. (I believe you need to supply the content of the card, and some way of signaling when it is supposed to be shown. There is probably just some interface that you have to implement.)
If there is no documented solution, a hackish/undocumented way would be ok, too. I'm mostly curious how it works.
Edit: Specifically, does somebody have knowledge about the internals of Google Now, e.g. by decompiling the .apk? What I've seen suggests it is pretty modular, and it should be fairly easy to drop another class into the .apk, or to maybe inject code using Cydia Substrate. I know that there is (as of Nov. 2013) no official way to add new cards.
There is currently no way to do that. Google makes its own cards and custom application cannot register any cards. But I hope it will be possible in future.
Actually Google announced last week that developers can now develop custom Google Now cards:
http://www.google.com/landing/now/integrations.html
However, a developer guide seems not available yet.
Edit:
On the end of the page they point out that:
We'll let you know when we are able to onboard more partners
There is a work-around that will soon allow you to place cards in Google Now's stream at a particular time or a particular location: Use Google Keep (https://drive.google.com/keep/)
You can create a new card at Google Keep with a time based or location based reminder, depending on which the relevant card will show up in Google Now.
Since Google Keep is now in Drive, the API is expected to be available soon (keep a lookout for it at http://discovery-check.appspot.com/ )
There is not way to do this by your own at the moment. If you really want to do it you can fill in this form: https://services.google.com/fb/forms/nowintegrations/. You can ask Google if they want to cooperate to create a Google Now card.
Note quite an answer, as it is still not possible to create Google Now cards, but you can now hook into the Google Now search function (basically Android's Siri) and provide custom search results. For example you can say "show me the lyrics to..." and it opens a lyrics app.
Here is a link to the project which is based on the Xposed framework.
Just guessing from my impression of the Google Search apk (which includes all the Google Now functionality and even the home screen on KitKat), it should be possible to use a similar technique to inject cards into the app - however since the app is huge and very complicated, it will be a lot of work. I'd keep my eyes open on the xda-developers forums, wouldn't be surprized if someone there solves this in the future.
It appears that there is developer documentation on how to push google now info via email, eg. flight details, restaurant reservations etc.
https://developers.google.com/schemas/now/cards
I have yet to dig into this, but may update this answer if I discover anything significant.
Say I have a set of say 100 stored images (say sports team logos, rather than very similar images like faces) in my android app. In a manner similar to google goggle's continuous mode, I would like to use the camera to decide which image is being focused on.
What is the best, most efficient way to accomplish this? Any open source libraries, sdks etc would be great.
Thanks.
If you would like serverside recognition you should check out kooaba's API. its free for small reference databases such as yours and the best in the field...
Www.kooaba.com
First, sorry abour my poor english.
I'm planning to build an augmented reality app for android mobile platform and the main feature is the ability of the user to take a shoot of a shop and the application recognize the shop that he is photographing. I Do not know if the best option would be to use an image recognition api as many existing, but I think it would be something more specific. Maybe own a bank of images would help.
My plan was to have a database of stores with their locations and use one of many tools for image recognition and search in my database to the same location. But I found that all search engines images (kooba, iqengines, etc.) are not free and not a little cheaper. So would a tool that could use a limited catalog, like shops images in a shopping mall for example and send photos of smartphones (both android or iphone).
Can someone help me get started?
I've been doing something similar for my dissertation at University. I developed an application which detected signposts, read the content on them, then personalised / prioritised it depending on the user's preferences (with mixed success).
As part of this I had to look into Image Recognition.
Two things you may want to look at are:
The Qualcomm QCAR SDK. This was a little bit too image specific for what I was after, but if you were to do it on a small range of shops it may work. This would require a collection of shop images to match against - I don't know how successful it would be.
What I implemented used JavaCV (a conversion of OpenCV), which also has an Android conversion. It seems to allow for image recognition a bit more generally than the previous option which is why I used it. It would require you to run your own training to create a classifier though (unless there is another way of doing image recognition within it). But there are a number of guides which can help with that.
I used it for recognising signposts with reasonable success off just some basic training, though did tend to recognise a number of false positives.
Within my application I then used location to match up with previous detections etc.
Hopefully these may get you started.