Ojos tackles tagging automation
EarlyStageVC: Deven->Ojos->?
Ojos, a new startup founded by Munjal Shah, the former CEO of Andale has gotten funded to pursue the illusive goal of auto-tagging of digital images. Coverage in Business Week Online.
This is a nice meaty problem to solve but it remains to be seen what the right combination of technologies and user interfaces will succeed in creating something that is really useful. I too have over 20,000 images on my hard disk and would like a better way to get them tagged and categorized. My experience with computer vision-based approaches for face recognition leads me to believe that coming up with a solution to
tagging the archive is a very, very difficult one (illumination & pose are tough nuts to crack).
That said, a mechanism where you could get assitance in tagging a set of images might be more reasonable. If I've taken 30 pictures at a birthday party and I identify all of the people as I go through the first 5 pictures, software should be able to help me identify those same people in the next 25. There will also be more and more intelligence getting built into the capture devices that will make the manual tagging chore easier over time. This intelligence ranges from the device being able to know its location when the picture is taken to information that can be set in the device (or on the web and communicated to the device) that would then be associated with a set of images shot during a specific time frame. This might be the 80% solution that will be good enough for most users.
This is certainly another interesting area and I'll be keeping an eye on Ojos as they make their products available to try.
Ojos, a new startup founded by Munjal Shah, the former CEO of Andale has gotten funded to pursue the illusive goal of auto-tagging of digital images. Coverage in Business Week Online.
This is a nice meaty problem to solve but it remains to be seen what the right combination of technologies and user interfaces will succeed in creating something that is really useful. I too have over 20,000 images on my hard disk and would like a better way to get them tagged and categorized. My experience with computer vision-based approaches for face recognition leads me to believe that coming up with a solution to
tagging the archive is a very, very difficult one (illumination & pose are tough nuts to crack).
That said, a mechanism where you could get assitance in tagging a set of images might be more reasonable. If I've taken 30 pictures at a birthday party and I identify all of the people as I go through the first 5 pictures, software should be able to help me identify those same people in the next 25. There will also be more and more intelligence getting built into the capture devices that will make the manual tagging chore easier over time. This intelligence ranges from the device being able to know its location when the picture is taken to information that can be set in the device (or on the web and communicated to the device) that would then be associated with a set of images shot during a specific time frame. This might be the 80% solution that will be good enough for most users.
This is certainly another interesting area and I'll be keeping an eye on Ojos as they make their products available to try.
posted by MyronK @ 1:51 PM
0 comments
