n_cols=instances,
scale=sigma,
make_plots=False,
edit_type=edit_mode)
plt.close('all')
frames = [x for _, x in frames]
frames = frames + frames[::-1]
make_mp4(
frames, 5, outdir_comp /
f'{get_edit_name(edit_mode)}_sigma{sigma}_comp{c}.mp4')
# Summary grid, random directions
# Using the stdevs of the principal components for same norm
random_dirs_act = torch.from_numpy(
get_random_dirs(n_comp, np.prod(sample_shape)).reshape(
-1, *sample_shape)).to(device)
random_dirs_z = torch.from_numpy(
get_random_dirs(n_comp, np.prod(inst.input_shape)).reshape(
-1, *latent_shape)).to(device)
for edit_mode in edit_modes:
plt.figure(figsize=(14, 12))
plt.suptitle(
f"{model.name} - {layer_name}, random directions w/ PC stdevs, {get_edit_name(edit_mode)} edit",
size=16)
make_grid(tensors.Z_global_mean,
tensors.Z_global_mean,
random_dirs_z,
tensors.Z_stdev,
tensors.X_global_mean,
random_dirs_act,
# One reasonable, one over the top
for sigma in [args.sigma, 3*args.sigma]:
for c in range(components):
for edit_mode in edit_modes:
frames = make_grid(tensors.Z_global_mean, tensors.Z_global_mean, tensors.Z_comp[c:c+1, :, :], tensors.Z_stdev[c:c+1], tensors.X_global_mean,
tensors.X_comp[c:c+1, :, :], tensors.X_stdev[c:c+1], n_rows=1, n_cols=instances, scale=sigma, make_plots=False, edit_type=edit_mode)
plt.close('all')
frames = [x for _, x in frames]
frames = frames + frames[::-1]
make_mp4(frames, 5, outdir_comp / f'{get_edit_name(edit_mode)}_sigma{sigma}_comp{c}.mp4')
# Summary grid, random directions
# Using the stdevs of the principal components for same norm
random_dirs_act = torch.from_numpy(get_random_dirs(n_comp, np.prod(sample_shape)).reshape(-1, *sample_shape)).to(device)
random_dirs_z = torch.from_numpy(get_random_dirs(n_comp, np.prod(inst.input_shape)).reshape(-1, *latent_shape)).to(device)
for edit_mode in edit_modes:
plt.figure(figsize = (14,12))
plt.suptitle(f"{model.name} - {layer_name}, random directions w/ PC stdevs, {get_edit_name(edit_mode)} edit", size=16)
make_grid(tensors.Z_global_mean, tensors.Z_global_mean, random_dirs_z, tensors.Z_stdev,
tensors.X_global_mean, random_dirs_act, tensors.X_stdev, scale=args.sigma, edit_type=edit_mode, n_rows=14)
plt.savefig(outdir_summ / f'random_dirs_{get_edit_name(edit_mode)}.jpg', dpi=300)
show()
# Random instances w/ components added
n_random_imgs = 10
latents = model.sample_latent(n_samples=n_random_imgs)
for img_idx in trange(n_random_imgs, desc='Random images', ascii=True):