No doubt you have seen the amazing capabilities of computer programs nowadays to recognize human faces, identify landmarks, and search similar images. This is all within the purview of computer vision, a field in computer science which seeks to analyze images and videos to duplicate human vision. It’s like making computers “see” in human terms, with our biological processes being represented in computer by high dimensional data.
For this post, I’m sharing my presentation in our computer vision class. We are given two papers with similar topics and we are to discuss them in 15 minutes. My topic was scene recognition. It’s quite impossible to talk about each paper in depth so we kept only the most salient points of each.
The first paper presented was “What Makes Paris Look Like Paris?”[1] and the second paper was “80 Million Tiny Images: A Large Dataset for non-parametric object and scene recognition” [2]. Both papers were very interesting. The first one described an algorithm that can recognize the “stylistic narrative” of a place, particularly Paris, while the second one discussed their huge dataset of tiny images and how computer scientists can use the data for scene recognition, person detection and image annotation.
Check out the links below for the project pages.
[1] Carl Doersch, Saurabh Singh, Abhinav Gupta, Josef Sivic, and Alexei A. Efros. What Makes Paris Look like Paris? ACM Transactions on Graphics (SIGGRAPH 2012), August 2012, vol. 31, No. 3. (http://graphics.cs.cmu.edu/projects/whatMakesParis/)
[2] Antonio Torralba, Rob Fergus, and William T. Freeman. 2008. 80 Million Tiny Images: A Large Data Set for Nonparametric Object and Scene Recognition. IEEE Trans. Pattern Anal. Mach. Intell. 30, 11 (November 2008), 1958-1970. DOI=10.1109/TPAMI.2008.128 (http://groups.csail.mit.edu/vision/TinyImages/)
It is the REAL DYL 