Lab 8    February 29, 2016  NAME ________________________ 

advertisement
CSCI 512 / EENG 512
Computer Vision
Spring 2016 Lab 8 NAME ________________________ February 29, 2016 NAME ________________________ This lab assignment should be done in teams of two. Go through the exercises below and show me your results. If you don’t finish by the end of today’s class, you may show me during the next class. Using the code developed in class, match the object “book1.pgm” to each of the images “Img0[i].pgm” and “TestImg[i].pgm” where [i]=1...10. Use N=4 for the threshold of the number of feature matches that determines whether the object is present in an image. In other words, if you get N or more matches, you say that the object is present in the image; otherwise, if you get fewer than N matches, you say that the object is not present. For the other algorithm parameters, use the values in the table below. Parameter Peak_thresh, Edge_thresh matching threshold for vl_ubcmatch Resolution of Hough parameter array Value Peak_thresh = 0 (0 = default) Edge_thresh = 10 (10 = default) Threshold = 2.0 (default is 1.5) x,y is divided into 5 bins theta is divided into 9 bins scale is divided into 5 bins Description Limits the number of features detected by ​
vl_sift Descriptor D1 is matched to a descriptor D2 only if the distance d(D1,D2) multiplied by thresh is not greater than the distance of D1 to all other descriptors This is the resolution of the Hough array that accumulates votes for a particular (x,y,theta,scale) In each of the 20 trials, note if this recognition method decides that the object is present. Also note (by looking at the images) whether the object was actually present in the image. Determine: True positive rate: ____________ False positive rate: ____________ Note: A “receiver operating curve” (ROC) is a useful analysis tool to set a threshold that optimizes the true positive rate against the false positive rate (see the Szeliski book, Figure 4.23). 1 
Download