Testing recognition with cropped class images


Procedure



1. Use the cropped, non-smiling students (in class_nonsmiling_cropped) to compute 10 eigenfaces. Show the average face and eigenfaces.


2. Use the same set of images to compute a userbase.
3. Have the program recognize the cropped, smiling students. You should expect only about 74% (20/27) accuracy with 10 eigenfaces. 
# of eigenFaces : # correct (correct matchings) 
1 : 3 (6, 9, 21)
3 : 12 (3, 4, 6, 8, 9, 11, 12, 14, 16, 19, 22, 24)
7 : 14 (2, 3, 5, 6, 8, 11, 12, 15, 16, 19, 22, 25, 27)
9 : 17 (2, 3, 5, 6, 8, 11, 12, 15, 16, 17, 18, 19, 22, 24, 25, 26, 27)
10 : 20 (2, 3, 5, 6, 7, 8, 11, 12, 14, 15, 16, 17, 18, 19, 22, 23, 24, 25, 26, 27)
11 : 21 (2, 3, 5, 6, 7, 8, 10, 11, 12, 14, 15, 16, 17 18, 19 , 22, 23, 24, 25, 26, 27)
13 : 20 (2, 3, 5, 6, 7, 8, 10, 11, 12, 14, 16, 17, 18, 19 ,22, 23, 24, 26, 27)
15 : 20 (2, 3, 5, 6, 7, 8, 10, 11, 12, 14, 16, 17, 18, 19, 22, 23, 24, 25, 26, 27)
17 : 21 (2, 3, 4, 5, 6, 7, 8, 10, 11, 12, 14, 16, 17, 18, 19, 22, 23, 24, 25, 26, 27)
19 : 20 (2, 3, 5, 6, 7, 8, 10, 11, 12, 14, 16, 17, 18, 19, 22, 23, 24, 25, 26, 27)
21 : 20 (2, 3, 5, 6, 7, 8, 10, 11, 12, 14, 16, 17, 18, 19, 22, 23, 24, 25, 26, 27)
23 : 20 (2, 3, 5, 6, 7, 8, 10, 11, 12, 14, 16, 17, 18, 19, 22, 23, 24, 25, 26, 27)
25 : 20 (2, 3, 5, 6, 7, 8, 10, 11, 12, 14, 16, 17, 18, 19, 22, 23, 24, 25, 26, 27)
27 : 20 (2, 3, 5, 6, 7, 8, 10, 11, 12, 14, 16, 17, 18, 19, 22, 23, 24, 25, 26, 27)


Questions

1. Describe the trends you see in your plots. Discuss the tradeoffs; how many eigenfaces should one use? Is there a clear answer?
The results improved drastically when moving from too few towards the "right" number of images. It appears that too many vectors does not hurt it that much though since the results did not drop off. It appears that the best results, for this batch, came from having either 11 or 17 eigenfaces. These numbers hover around half of the total faces to be comparing against, so I'm wondering if there is a relationship between the values.

2. You likely saw some recognition errors in step 3; show images of a couple. How reasonable were the mistakes? Did the correct answer at least appear highly in the sorted results?
12 vs 21
As you can see, this is a very understandable mistake to have made. It did appear highly, and it is easy to see why these mistakes were made. It's interesting that 12's smiling face could be paired to his non-smiling one correctly, but apparently 21's couldn't.


Cropping and finding faces


Procedure



1. Use the 10 eigenfaces file computed in the previous problem.
2. Use your program to crop the elf.tga image. Use min_scale,max_scale, step parameters of .45, .55, .01. You should be able to successfully

3. Find a digital picture of yourself; if you really don't have one, use any portrait on the web. Use your program to crop the picture.
4. Experiment with min_scale, max_scale, and step parameters to find ones that work robustly and accurately.
5. Find the faces in two different photos (use the crop=false option)


Questions

1. What min_scale, max_scale, and scale step did you use for each image?
For both of them : from .3 to .5 by .1

2. Did your attempt to find faces result in any false positives and/or false negatives? Discuss each mistake, and why you think they might have occurred.
Lots of false positives. Not so much as false negatives, just too many false positives that out-ranked the actual faces. I didn't factor in color due to time and energy not being available. These mistakes occured because the varriation obviously resembled a face. I can easily understand why fabric kept picking up as a result, it can often have folds the resemble eye sockets, mouth, and nose lines.