Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Ease the visualization of latent directions #120

Open
wants to merge 3 commits into
base: master
Choose a base branch
from
Open

Ease the visualization of latent directions #120

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
6930884
Add CLI argument to sample more scales for moving in a direction
woctezuma Sep 19, 2020
25b87a0
Transpose the grid of results (one sample per row)
woctezuma Sep 19, 2020
5fa9f30
Fix typo
woctezuma Sep 19, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
55 changes: 34 additions & 21 deletions apply_factor.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,6 +27,13 @@
default=2,
help='channel multiplier factor. config-f = 2, else = 1',
)
parser.add_argument(
"-d_num",
"--degree_num",
type=int,
default=3,
help="number of scalar factors for moving latent vectors along eigenvector",
)
parser.add_argument("--ckpt", type=str, required=True, help="stylegan2 checkpoints")
parser.add_argument(
"--size", type=int, default=256, help="output image size of the generator"
Expand Down Expand Up @@ -64,31 +71,37 @@
latent = torch.randn(args.n_sample, 512, device=args.device)
latent = g.get_latent(latent)

direction = args.degree * eigvec[:, args.index].unsqueeze(0)
direction = eigvec[:, args.index].unsqueeze(0)

img, _ = g(
[latent],
truncation=args.truncation,
truncation_latent=trunc,
input_is_latent=True,
)
img1, _ = g(
[latent + direction],
truncation=args.truncation,
truncation_latent=trunc,
input_is_latent=True,
)
img2, _ = g(
[latent - direction],
truncation=args.truncation,
truncation_latent=trunc,
input_is_latent=True,
)
img_dict = dict()

for u in torch.linspace(- args.degree, args.degree, args.degree_num):

img_batch, _ = g(
[latent + u * direction],
truncation=args.truncation,
truncation_latent=trunc,
input_is_latent=True,
)

for j in range(img_batch.shape[0]):

img = img_batch[j].unsqueeze(0)

try:
img_dict[j].append(img)
except KeyError:
img_dict[j] = [img]

img_list = [
torch.cat(img_dict[j], 0)
for j in range(args.n_sample)
]

grid = utils.save_image(
torch.cat([img1, img, img2], 0),
torch.cat(img_list, 0),
f"{args.out_prefix}_index-{args.index}_degree-{args.degree}.png",
normalize=True,
range=(-1, 1),
nrow=args.n_sample,
nrow=args.degree_num,
)