regions = np.array(bb_to_seg(gen_bbs))[0, :, :, :].transpose((1, 2, 0))
graph = [real_nodes, None]
if k == 0:
graph_arr = draw_graph([real_nodes, eds.detach().cpu().numpy()])
final_images.append(graph_arr)
# # place real
# real_bbs = real_bbs[np.newaxis, :, :]/32.0
# real_im = bb_to_im_fid(real_bbs, real_nodes)
# final_images.append(real_im)
# reconstruct
fake_im_seg = draw_masks(gen_mks, real_nodes)
final_images.append(fake_im_seg)
fake_im_bb = bb_to_im_fid(gen_bbs, real_nodes,
im_size=256).convert('RGBA')
final_images.append(fake_im_bb)
n_rows += 1
if (n_rows + 1) % 12 == 0:
final_images_new = []
for im in final_images:
print(np.array(im).shape)
final_images_new.append(
torch.tensor(np.array(im).transpose((2, 0, 1))) / 255.0)
# print('final: ', final_images_new[0].shape)
final_images = final_images_new
final_images = torch.stack(final_images)
print(final_images)
save_image(final_images,
"./output/results_page_{}_{}.png".format(
[np.array(mask_to_bb(mk)) for mk in gen_mks.detach().cpu()])
real_bbs = np.array(
[np.array(mask_to_bb(mk)) for mk in real_mks.detach().cpu()])
real_nodes = np.where(given_nds.detach().cpu() == 1)[-1]
gen_bbs = gen_bbs[np.newaxis, :, :] / 32.0
junctions = np.array(bb_to_vec(gen_bbs))[0, :, :]
regions = np.array(bb_to_seg(gen_bbs))[0, :, :, :].transpose((1, 2, 0))
graph = [real_nodes, None]
if k == 0:
graph_arr = draw_graph([real_nodes, eds.detach().cpu().numpy()])
final_images.append(graph_arr)
# place real
real_bbs = real_bbs[np.newaxis, :, :] / 32.0
real_im = bb_to_im_fid(real_bbs, real_nodes)
final_images.append(real_im)
# reconstruct
fake_im = draw_masks(gen_mks, real_nodes)
# fake_im = bb_to_im_fid(gen_bbs, real_nodes)
final_images.append(fake_im)
row = 0
for k, im in enumerate(final_images):
path = './out/var_{}/'.format(row)
os.makedirs(path, exist_ok=True)
im.save('{}/{}.png'.format(path, k))
if (k + 1) % 12 == 0:
row += 1
# final_images = torch.stack(final_images).transpose(1, 3)
# Unpack batch
mks, nds, eds, nd_to_sample, ed_to_sample = batch
# Configure input
real_mks = Variable(mks.type(Tensor))
given_nds = Variable(nds.type(Tensor))
given_eds = eds
for k in range(opt.num_variations):
# plot images
z = Variable(
Tensor(np.random.normal(0, 1,
(real_mks.shape[0], opt.latent_dim))))
with torch.no_grad():
gen_mks = generator(z, given_nds, given_eds.cuda())
gen_bbs = np.array(
[np.array(mask_to_bb(mk)) for mk in gen_mks.detach().cpu()])
real_bbs = np.array(
[np.array(mask_to_bb(mk)) for mk in real_mks.detach().cpu()])
real_nodes = np.where(given_nds.detach().cpu() == 1)[-1]
if k == 0:
real_bbs = real_bbs[np.newaxis, :, :] / 32.0
real_im = bb_to_im_fid(real_bbs, real_nodes)
real_im.save('{}/{}.jpg'.format(path_real, globalIndexReal))
globalIndexReal += 1
# draw vector
gen_bbs = gen_bbs[np.newaxis, :, :] / 32.0
fake_im = bb_to_im_fid(gen_bbs, real_nodes)
fake_im.save('{}/{}.jpg'.format(path_fake, globalIndexFake))
globalIndexFake += 1