Everyone’s talking about the new Riya photo technology with face recognition, and I’ve been lucky enough to wangle my way onto the Alpha preview of the technology. I thought you might like to hear my first impressions, which are….it works! Oh yes, it may be ‘so AlphaItHurts’, and somewhat sluggish round the edges, but it really is a clever bit of coding no question. Read on for more early impressions…
The basic premise of Riya is that technology can remove the dreadful drudge of having to name your photos, and laboriously catalogue them so that you can find them later. The idea is that you upload your photographs, give names to the faces you want to identify in the future (Riya recommends a minimum of 10 shots of a particular person should be identified) and the system will recognise and automatically identify and tag with the right name all photos that you upload from that moment on.
Sound ambitious enough for you? The wonder is that it works as advertised. The first thing you need to do is download a small Uploader program, which lets you browse your hard disk and select folders or images to upload. The utility is sparse but functional. Give it your username and password, select a folder and press upload and leave it be. You can also set it to automatically upload images as you add them to your computer, which will clearly be a major use of the system eventually (it’s a bit slow for that at the moment I suspect).
I uploaded a small folder’s worth of images last night – they recommend that because the Alpha system is CPU intensive (read slow, as in 1 GB of images in 4 hours) you do it overnight – and this morning spent 10 minutes manually tagging around 14 of the 46 photos I’d uploaded. The system then went ahead and correctly located and identified the remaining images with no fuss whatsoever. This part doesn’t seem to be that slow, the 20 or so images were tagged within a few minutes, so larger collections shouldn’t take that long, and let’s face it, manual tagging is not exactly a quick process is it?
Another clever fact is that the system will also hunt for other images with the same identification (i.e. name) and add them to the bottom of your search page of images, even if they’re in someone else’s online collection. This means that several family members could store and share their images really easily without bothering to catalogue or even name the photos – at last, an end to the CIMG3838.jpg tyranny. Oh, and did I mention that there is text recognition included in the mix somewhere to add even more to the searching functionality? So for instance if you go to a particular area all you have to do is shoot the hotel sign and all your images will be automatically tagged and associated with that trip. The engine will look at all the text in your images and assign them as search keywords. Yikes!
There are lots of other features planned or included in this technology preview, like email tagging to make this sharing of images between friends and family – or even the public at large – even easier (no need to worry about privacy as long as the photo has the right email address attached – see below), and the company is clearly working flat out to deliver a whole raft of new applications for when the product finally reaches a commercial state, but for now it’s probably just fair to say that the technology works, it’s not hype, and this really could be a huge, huge development in the digital photography marketplace in general. For more of the Riya flavour check out the CEO’s frank and very genuine weblog and the official product blog.
So just to round out, here are a few screenshots over the page to give a flavour of the interface (click on thumbnails)