Edge of Vision

After a hiatus of more than a year, I’ve decided to try reactivating this jolly ol’ blog. As before, my plan is to periodically share commentary on various interesting happenings in neuroscience and artificial intelligence.

As for myself, for the past year I’ve been doing research using transcranial magnetic stimulation (TMS) to reactivate recently-viewed visual percepts in occipital cortex. The basis of the work can be found in the PhD thesis of my colleague Daw-An Wu, titled “How perception adheres color to objects and surfaces : studies using visual illusions and transcranial magnetic stimulation.” At some point I’ll elaborate on the directions we’ve been pursuing in the past year, which have had some rather fascinating results.

Some time ago I realized that I couldn’t find any good sites which discuss recent research in computer vision. This “research blog” is intended to help fill that gap. My plan is to discuss recent and/or important research papers about computer vision and biological vision, news articles which involve vision, and vision-related ideas I have which may be interesting/useful. I’m also a big fan of the related topics of neuroscience and robotics, so I’ll probably regularly post about those as well.Although vision research can be a pretty technical at times, I’m hoping that at least some of this blog will be interesting and understandable by those without prior experience in the field.
A few things I hope to post about in the near future:

• Luis von Ahn’s work with using web-based games to extract huge amounts of visual knowlede from humans.
• Riya, a commercial web-based product for face recognition in galleries of photographs.
• The recent PhD thesis defense of fellow CNS grad student Dirk Walther, involving the use of visual attention models to facilitate learning of models for object recognition.
• The landmark Viola & Jones algorithm for face detection, its extensions, and the excellent open-source implementation of it in Intel’s OpenCV toolkit.
• My thoughts on various feature descriptors useful in computational object recognition, such as David Lowe’s SIFT and Alex Berg’s geometric blur.