Face Recognition: ItemList on the y axis is always zero

Hey everybody,

I started writing my own face recognition. My dataset consists of images with faces which are labeled with two coordinate points to describe the position of a rectangle around them.

I was able to load the images into an ImageList and also add the coordinates as labels by writing my own labeling function.

My problem is that I get an error when calling the show_batch function afterwards:

AttributeError: 'NoneType' object has no attribute 'reconstruct'

Also the output that’s printed out after calling label_from_func() seems wrong to me:


Train: LabelList (328 items)
x: ImageList
Image (3, 333, 650),Image (3, 697, 1280),Image (3, 240, 460),Image (3, 291, 600),Image (3, 500, 500)
y: ItemList
0,0,0,0,0
Path: .;

Valid: LabelList (81 items)
x: ImageList
Image (3, 472, 630),Image (3, 438, 779),Image (3, 366, 550),Image (3, 730, 1296),Image (3, 448, 495)
y: ItemList
0,0,0,0,0
Path: .;

Test: None

The ImageList on the x axis seems to be fine, but the ItemList on the y axis seems strange to me. Why is it only filled with zeroes?

This is how my labels look like:

[(0.08615384615384601, 0.306306306306306),
 (0.17230769230769202, 0.45345345345345306)]

Any help is much appreciated.

Kind regards from Germany

Okay, it turned out that my labeling function had a bug. Everything’s working fine now.