Image Recognition and the Future

My colleague over at PDNPulse - Daryl Lang - wrote about Google (Google: We Will Organize You, 7/1/08) , and the concept of autonomously recognizing subject matter within an image.

This reminded me of conference I attended - Department of Energy Computational Science Annual Fellows' Conference, where I was fascinated by a presentation by Kristen Grauman on Scalable Image Recognition and Retrieval. Grauman was the recipient of a scholarship for her work in this field, and her presentation - all 43 minutes of it, can be seen here (stick with it for the first 45 seconds of audio, her visual presentation starts at 45 seconds). Having seen it live, I still enjoy watching, and re-watching, the presentation online.

If you'd like to see even more of her research, including her 2006 presentation that I also saw, click here. For those of you interested in this, it's remarkable to see how concepts like human to computer interfaces, manual keywording, and image sorting by subject will be done much more automatically in the future. This also may be one of the ways in which images are recognized in the Orphan Works issue.

Image a photograph of a panda bear, with a link next to it that reads "Click to view more like this". Kristen, and Google, will likely be among the solutions to the visual database. At the 3:13 mark in the first video Kristen talks about object recognition within an image (identifiying a specific person or building), and the category mode (like buildings being similar, or animals, and so forth). It REALLY starts to get interesting at about 6 minutes, but BOTH videos are well worth your time. (Check the PDF's of each presentation that are available via the link for a higher resolution version of each presentation, and the abstract above the video.) Her bio as a Clare Boothe Luce Assistant Professor, Department of Computer Sciences, University of Texas at Austin, can be seen here, which includes her e-mail address if you're so inclined to learn more about her research.

Google's R.J. Pittman, Director of Product Management (as seen on Beet.tv here) should be calling Kristen ASAP about her research!

Fascinating.

(Comments, if any, after the Jump)

---

Please post your comments by clicking the link below. If you've got questions, please pose them in our Photo Business Forum Flickr Group Discussion Threads.