def test_V3Model():
from data.data_loader import CreateDataLoader
from options.multimodal_gan_options_v3 import TrainMMGANOptions_V3 as Option
from models.multimodal_designer_gan_model_v3 import MultimodalDesignerGAN_V3 as Model
from misc.visualizer import GANVisualizer_V3
opt = Option().parse(
'--debug --batch_size 1 --nThreads 1 --shape_encode flx_seg --shape_with_face 1'
)
data = iter(CreateDataLoader(opt)).next()
visualizer = GANVisualizer_V3(opt)
model = Model()
model.initialize(opt)
model.set_input(data)
model.forward()
for k, v in model.output.iteritems():
try:
print('%s: %s' % (k, v.size()))
except:
print('%s' % k)
visuals = model.get_current_visuals(mode='input')
visualizer.visualize_image(1, subset='input', visuals=visuals)
############################################
opt = TestPoseParsingOptions().parse()
train_opt = io.load_json(os.path.join('checkpoints', opt.id, 'train_opt.json'))
preserved_opt = {'gpu_ids', 'is_train'}
for k, v in train_opt.iteritems():
if k in opt and (k not in preserved_opt):
setattr(opt, k, v)
# create model
model = PoseParsingModel()
model.initialize(opt)
# save terminal order line
io.save_str_list([' '.join(sys.argv)], os.path.join(model.save_dir, 'order_line.txt'))
# create data loader
val_loader = CreateDataLoader(opt, split='test')
# create visualizer
visualizer = GANVisualizer_V3(opt)
############################################
# Visualize
############################################
if opt.nvis > 0:
print('visualizing first %d samples' % opt.nvis)
num_vis_batch = int(np.ceil(1.0*opt.nvis/opt.batch_size))
visuals = None
for i, data in enumerate(val_loader):
if i == num_vis_batch:
break
model.set_input(data)
model.test(compute_loss=False)
v = model.get_current_visuals()
if visuals is None:
def compare_segment():
'''
Compare segment result of two methods:
1. segment model used in FashionGAN (Be Your Own Prada: Fashion Synthesis with Structural Coherence, ICCV 2017)
2. segment model used in VITON (VITON: An Image-based Virtual Try-on Network, CVPR 2018)
- Look into Person: Joint Body Parsing & Pose Estimation Network and A new Benchmark, T-PAMI 2018
'''
import torch
import torchvision
from misc.visualizer import GANVisualizer_V3
pair_list = io.load_json(
'datasets/DF_Pose/Label/pair_split.json')['test'][0:64]
img_dir = 'datasets/DF_Pose/Img/img_df/'
seg_atr_dir = 'datasets/DF_Pose/Img/seg_df/'
seg_lip_dir = 'datasets/DF_Pose/Img/seg-lip_df/'
seg_lip_rv_dir = 'datasets/DF_Pose/Img/seg-lip_df_revised/'
visuals = defaultdict(lambda: [])
for (id_1, id_2) in pair_list:
visuals['img_1'].append(imageio.imread(img_dir + id_1 + '.jpg'))
visuals['seg_atr_1'].append(imageio.imread(seg_atr_dir + id_1 +
'.bmp'))
visuals['seg_lip_1'].append(imageio.imread(seg_lip_dir + id_1 +
'.bmp'))
visuals['seg_lip_rv_1'].append(
imageio.imread(seg_lip_rv_dir + id_1 + '.bmp'))
visuals['img_2'].append(imageio.imread(img_dir + id_2 + '.jpg'))
visuals['seg_atr_2'].append(imageio.imread(seg_atr_dir + id_2 +
'.bmp'))
visuals['seg_lip_2'].append(imageio.imread(seg_lip_dir + id_2 +
'.bmp'))
visuals['seg_lip_rv_2'].append(
imageio.imread(seg_lip_rv_dir + id_2 + '.bmp'))
visuals = {k: np.stack(v, axis=0) for k, v in visuals.iteritems()}
for name in ['img_1', 'img_2']:
visuals[name] = (torch.Tensor(visuals[name].transpose(
0, 3, 1, 2)).div_(127.5).sub_(1), 'rgb')
for name in [
'seg_atr_1', 'seg_atr_2', 'seg_lip_1', 'seg_lip_2', 'seg_lip_rv_1',
'seg_lip_rv_2'
]:
visuals[name] = (torch.Tensor(visuals[name][:,
np.newaxis, :, :]), 'seg')
visuals_ordered = OrderedDict()
for k in [
'img_1', 'seg_atr_1', 'seg_lip_1', 'seg_lip_rv_1', 'img_2',
'seg_atr_2', 'seg_lip_2', 'seg_lip_rv_2'
]:
visuals_ordered[k] = visuals[k]
imgs, vis_list = GANVisualizer_V3.merge_visual(visuals_ordered)
print(vis_list)
output_dir = 'temp/df_seg/'
io.mkdir_if_missing(output_dir)
torchvision.utils.save_image(imgs,
output_dir +
'compare_segment_atr-and-lip.jpg',
nrow=8,
normalize=True)