Given how many photos we’re taking these days, organising them and accessing them still seems an unsolved problem. Apple’s taking its stab at the issue, but we’re still faced with huge organisational challenges – that most people never bother with. More photos are vanishing into rarely examined archives than ever before.

We need help. Computational help. But computers are lousy at recognising images.

Well, it appears we’re getting ever closer to solving that problem. In fact, Microsoft’s researchers have just passed a major image recognition milestone:

the researchers say their system achieved a 4.94 percent error rate on the 1000-class ImageNet 2012 classification dataset, which contains about 1.2 million training images, 50,000 validation images, and 100,000 test images. In previous experiments, humans have achieved an estimated 5.1 percent error rate.

Yes, their system is now better at recognising this in that set of reference images than humans are.

Of course, it’s better not to get too carried away with this. As the researchers themselves note:

While our algorithm produces a superior result on this particular dataset, this does not indicate that machine vision outperforms human vision on object recognition in general…On recognizing elementary object categories…machines still have obvious errors in cases that are trivial for humans. Nevertheless, we believe our results show the tremendous potential of machine algorithms to match human-level performance for many visual recognition tasks.

Still, this is news that brings some potential with it. We’re generating ever larger numbers of photos tan ever before – and they’re probably less organised, sortable and searchable than they’ve ever been, too.

Every step towards computers being able to recognise the content of those images, tag them, and organise them for us, seems welcome.

[via Fast Company]

Photo by Susanne Nilsson and used under a Creative Commons licence