front_x or cross_x?

[sunnyHelen]

cross-view/crossView/CrossViewTransformer.py

Line 52 in 957691b

front_res = torch.cat((front_x, T), dim=1)

cross-view/crossView/CrossViewTransformer.py

Line 55 in 957691b

output = front_x + front_res

Hi, I was confused about the calculation of cross-view transformer. According to the Figure 4 in your paper, it seems that the "front_x" should be "cross_x" in Line 52 and 55. Am I misunderstanding anything? Please help me figure it out.
Looking forward to your reply and thanks!

[ChristopheZhao]

I have test the difference between front_x and cross_x that concanate with T:
sample 1(cross_x is better):
origin front:

the res used front_x:

the res used cross_x:

sample 2(cross_x is better):
origin front:

the res used front_x:

the res used corss_x:

sample 3(front_x is better):

the res used front_x:

the res used cross_x:

And more 10 sample i tested has same degree result that i cant distinguishing their difference,so i confused if the pretrain model used front_x concanated T to train data.But maybe the right ways is to concanated with cross_x,So although use cross_x concanated with T to test sample,also can achive the same result as font_x.