Class NoScale_RotatedFrontalFace_CDataGenerator

      ml.ml.DataGenerator --+    
                            |    
ml.cla.dataset.CDataGenerator --+
                                |
                               NoScale_RotatedFrontalFace_CDataGenerator

Generate face and non-face image patches of size 26-by-26.

Frontal faces can have:

• a : rotation angle in multiples of 45 degree, clockwise

  a = (M_PI / 4) * (i_a - 4), where i_a in {0..7}

• s : size 19x19 (fixed)

• x : shift in x in {-3.5, -2.5, ..., 2.5, 3.5} pixels

  x = i_x - 3.5, where i_x in {0..7}

• y : shift in y in {-3.5, -2.5, ..., 2.5, 3.5} pixels

  y = i_y - 3.5, where i_y in {0..7}

We have in total 8*8*8 = 512 face locations (a,x,y). Use static method class_to_location() to convert from class to location.

Correspondence between a class and a location

class j, Location a = (a2, a1, a0), y = (y2, y1, y0), x = (x2, x1, x0), then: j = (a2, a1, a0, y2, x2, y1, x1, y0, x0) + 1

Neighboring patches:

• Translation: move by 8 pixels.

Compared to traditional methods:

Shifting: traditional 1.5 pixel, ours 1.0 pixel