about embedding concat

[chengweige]

layer1 ~ layer3 have different feature map size, so how to concate embedding from layer1 ~ layer3?
i can't find it from the paper

[plutoyuxie]

layer1 ~ layer3 have different feature map size, so how to concate embedding from layer1 ~ layer3?
i can't find it from the paper

The concatenation function is defined here:

PaDiM-Anomaly-Detection-Localization-master/main.py

Line 286 in 65adb9c

def embedding_concat(x, y):

You could also use torch.nn.functional.interpolate(input, scale_factor=2, mode='nearest') as you like.

[Omarelsaadany]

Hello
Have any one got this error
x = x.view(B, C1, -1, H2, W2)
RuntimeError: shape '[20, 256, -1, 50, 50]' is invalid for input of size 200724480

[bjaeger1]

@Omarelsaadany. you need to change your input image size, as with your current input image size there is an dimension error. 200724480/(20x256x50x50) is a float number and not an integer
The dimension "-1" is a PyTorch alias for "infer this dimension given the others have all been specified" (i.e. the quotient of the original product by the new product)