RuntimeError: The size of tensor a (16) must match the size of tensor b (8) at non-singleton dimension 3

[G-force78]

RuntimeError Traceback (most recent call last)
in <cell line: 8>()
6 shape_path = '/content/HairFastGAN/womancropped512x512.png'
7 color_path = '/content/HairFastGAN/mr.png'
----> 8 final_image = hair_fast.swap(face_path, shape_path, color_path)
9 # T.functional.to_pil_image(final_image).resize((512, 512)) # 1024 -> 512
10 import torchvision.transforms as transforms

11 frames
/content/HairFastGAN/models/FeatureStyleEncoder/pixel2style2pixel/models/stylegan2/model.py in insert_feature(x, layer_idx)
527 def insert_feature(x, layer_idx):
528 if features_in is not None and features_in[layer_idx] is not None:
--> 529 x = (1 - feature_scale) * x + feature_scale * features_in[layer_idx].type_as(x)
530 return x
531

RuntimeError: The size of tensor a (16) must match the size of tensor b (8) at non-singleton dimension 3

[G-force78]

Input images need to be 1024x1024

[maximkm]

Input images need to be 1024x1024

Alternatively, you can simply use the 'align' flag. This flag, in addition to cropping images by face, will also resize the image to a resolution of 1024. Here's how you can use it:

final_image = hair_fast.swap(face_path, shape_path, color_path, align=True)