N=0
ngf = 80
ndf = 80
netMix =NetUskip(ngf, nDep, nz, bSkip=skipConnections, nc=3, ncIn=3, bTanh=True, Ubottleneck=Ubottleneck)
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
print ("device",device)
netMix.load_state_dict(torch.load(model_path))
netMix.to(device)
netMix.eval()
files = os.listdir(test_img_path)
for f in files:
if f == 'ying.jpg':
t0 = time.time()
print('picture_name:',f)
im = getImage(os.path.join(test_img_path,f),bDel = True)
im = im.to(device)
mask = getImage('samples/ying_maskk_dilated.png',bDel = True)
mask = mask.to(device)
im1=im
im = torch.mul(im,mask)
fixnoise2 = torch.FloatTensor(1,nz, im.shape[2] // 2 ** nDep, im.shape[3] // 2 ** nDep)
fixnoise2 = fixnoise2.to(device)
fixnoise2=setNoise(fixnoise2)
'''
else:
if False:
print('false')
drift=(fixnoise2*1.0).uniform_(-1, 1)
fakebig, alpha, A, mixedbig = famosGeneration(targetMosaic, fixnoise, templates.unsqueeze(0), True)
else:
fakebig, alpha, A, mixedbig = splitW(targetMosaic, fixnoise, templates.unsqueeze(0), famosGeneration)
vutils.save_image(mixedbig,'%s/mixed_epoch_%03d_%s.jpg' % (opt.outputFolder, epoch,desc), normalize =True)
if True:#
vutils.save_image(alpha,'%s/alpha_epoch_%03d_%s.jpg' % (opt.outputFolder, epoch,desc), normalize=False)
vutils.save_image(rgb_channels(A), '%s/blenda_epoch_%03d_%s.jpg' % (opt.outputFolder, epoch, desc), normalize=False)##already 01
v=nn.functional.avg_pool2d(A.view(-1, 1, A.shape[2], A.shape[3]), int(16))
vutils.save_image(v,'%s/blendaBW_epoch_%03d_%s.jpg' % (opt.outputFolder, epoch,desc), normalize=False)
vutils.save_image(fakebig,'%s/mosaicBig_epoch_%03d_%s.jpg' % (opt.outputFolder, epoch,desc),normalize=True)
##RUN OUT-OF-SAMPLE
with torch.no_grad():
try:
im=getImage(opt.testImage, bDel=True)
if im.shape[2]>targetMosaic.shape[2] or im.shape[3]>targetMosaic.shape[3]:
print ("cropping to original mosaic size")
im=im[:,:,:targetMosaic.shape[2],:targetMosaic.shape[3]]
im=torch.cat([im,targetMosaic[:,3:5,:im.shape[2],:im.shape[3]]],1)##coords
print ("test image size",im.shape)
fixnoise2 = torch.FloatTensor(1, nz, im.shape[2] // 2 ** nDep,im.shape[3] // 2 ** nDep)
fixnoise2 = fixnoise2.to(device)
fixnoise2 =setNoise(fixnoise2)
fakebig,_,_,_= splitW(im, fixnoise2, templates.unsqueeze(0), famosGeneration)
vutils.save_image(fakebig, '%s/mosaicTransfer_epoch_%03d_%s.jpg' % (opt.outputFolder, epoch, desc), normalize=True)
except Exception as e:
print ("test image error",e)
for net in Gnets:
net.train()