I got an email from the folks at Riya letting me know that the water, while very beta, is safe enough to venture into for face recognition. It's not like I'm an insider there - I just signed-up for the beta out of curiosity for a better way to organize my thousands of photos by person. So I installed the first beta and connected it to my computer. The first thing it asks you to do is to upload thousands of pictures. This is a fair request but it could use some enhancements on the specificity of the folders to upload. It basically by default will upload your My Pictures directory. Unfortunately not all of these pictures are meant for sharing and I believe I posted sub-sets of pictures from directories using Picasa. But I gave it 2005 to play with and I don't think anything terrible will happen as it goes about uploading the pictures.

The next step is the recognition. I like the recognition utility for manual tagging to train the engine. It crashed the first time I tried it with only two pictures but then when I tried it again with about eight pictures from the Superbowl party in 2005 it managed to find the individual folks who were there. I think that it did a good job of getting people separated out. But after training it on about six people and then submitting it the system went into a strange GIF image that was doing something for twenty minutes and appearing to show that something was happening recognizing faces. This was annoying because if something is going to take more than a few minutes then just let it happen in the background and then email me to login when it is done. I don't want to look for hours at some animated graphic that doesn't do anything waiting for it to prompt me to identify an ambiguous face. That's as far as I got for now because Riya needs volumes of images to recognize much and tag things automatically.

They suggest that 1000 images is a good minimum to work with and ADSL speeds lead me to believe that I'll be waiting about 60 hours to get that many images uploaded. Among my concerns with their upload tool is how it handles interrupts since not only does my DSL go slowly I believe it sometimes gets slow and I restart the router to resolve problems. In this case I'm not sure what the uploader will do.

I also noticed that they are planning on "exporting" the tags out to Flickr or "Desktop organizer". My hope is that the desktop one they are planning is Picasa although they didn't mention it. It would make sense for them to go for the two big players for interoperability. Else there would be nobody to acquire them. The integration ought to be interesting although I am still not sure about Flickr when it comes to thousands of photos. Don't they charge for posting photos. They had all sorts of limitations on volumes for "free" posting back when I first tried it so I didn't go back given that I know I'll be posting tons of pictures given the infancy ga-ga phase of Madeline and me. Maybe I'll check back with Flickr to see what their deal is with volumes of photos now that they are owned by Yahoo! I still am not sure why they don't just build the photo recognition and tagging tools into the desktop photo organizers (so I don't need to upload everything just to recognize things) but it could be a scale thing for them. I would prefer starting out to just install a Picasa plugin or something that runs on the desktop as an independent organizer utility that could let me do what the web application is doing without uploading all the images but uploading them doesn't stop me from wanting to test/use/play with the technology.