Closed Lotayou closed 9 months ago
Could it be that you are inputting a 4 dimensional tensor to the model? It seems pointless to have the shape of your input be (1, h*s, w*s, 25)
when (h*s, w*s, 25)
gives the same amount of information.
You could try doing input.squeeze()
before passing it into the model.
Here's the original piece of code
https://github.com/varunagrawal/tiny-faces-pytorch/blob/dbe0ee57ddc41891b5228ec244946b84480d87bf/models/utils.py#L36
However the actual tensor is 4-dimensional of shape![image](https://user-images.githubusercontent.com/33449901/84123313-43dd7c00-aa6c-11ea-8097-99b4c777d816.png)
(1,h*s,w*s,25)
, which would lead to an out-of-bound error when the input image resolution is lower than a certain threshold (900 I presume):I guess this line should be written as