Object Recognition For Photo Management

Organizing pictures is a terrible experience. The recent advances to photo organizing has been to add the “where,” “who” and “when” to our photos but sadly the “what” has been ignored. That is fine for the general public but there is a large demographic of artists and designers who could benefit from having their photos tagged with their contents. I have thousands of pictures saved from the web that I store in folder hierarchies with tags. I like to save pictures of great design for future inspiration. Millions of people save pictures from the web but there is no app to easily organize them. The current solutions are all manual, arduous tasks. Recently an app called Ember gave me the idea to fix this problem. Ember does something pretty cool for artists, it sorts pictures by colour. Sorting by colour is fairly useful but I see a much greater potential for this kind of organization, object recognition. Imagine this, you import a bunch of photos and they are analyzed for the objects they contain. When you want to pull up all of your architecture pictures you just press on the smart tag and you see all your pictures of buildings. To make this system even better I think it needs to take a page out of Pinterests book and be able to narrow down the results by style as well. So you could search for “architecture” and “modern” and get all your pictures of modern buildings. That is the picture management app that I want. You never have to spend hours renaming files, organizing pictures or tagging them. I think a smart move would be to allow users to favourite tags and just view the ones they want. This would combat the sheer amount of objects that their pictures contain. Just imagine a picture of a bathroom, it has a sink, bathtub, shower, soaps, towels, rug, etc. Most people wouldn’t want to search for bathtubs but the group of people that do, could. A system like this could adapt to the user very easily and for the creative types it could be an indispensable tool. Embers tagline is “save your inspiration" and it does that. The problem is finding it again means sorting through thousands of pictures or manually organizing everything yourself. Ember isn’t bad but it could be much, much better.

How Would This Work?

You upload a photo and object recognition is done on it. The picture is then tagged with every object the image contains. To find their pictures the user can either search, use their tags as a folder system, or narrow down the tags to find exactly what they want. The first two are pretty self explanatory but I think the third needs some clarification. Users can select multiple tags to narrow down results. They could press on “architecture” and “modern” to see all their modern architecture. To empower the user even more the system would allow them to use “and,” “or” and “not” in between tags. So “architecture” not “modern” would show all non-modern buildings. To always know where the stuff you need is you could add your own tags like “Design Folder” but this would be largely unnecessary.

This Would Probably Be Hard To Do

Object recognition is hard, according to just about everything I read online but is that any reason not to do something? A system like this would be such an invaluable tool for designers, artists, realtors, etc that I think the effort is worth it. It’s not like object recognition isn’t possible either. The Amazon Fire Phone does object recognition but the purpose is very different. Amazon lets you scan physical objects and then tries to sell them to you. This would analyze pictures to organize them.

Google and Stanford Just Made Some Huge Advances In Photo Recognition

After I had written the majority of this post I saw an article about Google and Stanford’s advancements in the area of object recognition. They were working on very similar projects, unbeknownst to each other. What they accomplished was software able to describe a picture with 60% accuracy. The NY Times had a good write up about it. Needless to say I am incredibly excited for what further developments these teams can create.

Final Words

Picture management has been stagnant for years. Any sort of photo management advances have mostly focused on where, when and who but they ignore the what. For the general public that is fine but there is a huge demographic of creatives out there who could benefit from adding the “what” to there pictures.


Agree? Disagree? Comment Below.

If you're interested in following my blog:

Subscribe via RSS

Follow me on Twitter: @aestusblog

Or subscribe by e-mail below.

Enter your email address: