def load_model(opt):
if opt.pretrained_file != "":
model = torch.load(opt.pretrained_file)
else:
if opt.model_def == 'gazenet':
model = gazenet.Net(opt)
if opt.cuda:
model = model.cuda()
return model
def load_model(opt):
model = gazenet.Net(opt)
return model.cuda()
from torchvision import transforms
import matplotlib.pyplot as plt
import matplotlib
parser = opts.optionargparser()
opt = parser.parse_args()
opt.testbatchsize = 16
checkpoint = torch.load(
'./savedmodels/gazenet_gazefollow_softmaxnesterovsgd_70epoch.pth.tar')
print("Loading pretrained model: ")
start_epoch = checkpoint['epoch']
best_err = checkpoint['best_err']
model = gazenet.Net(opt).cuda()
model.load_state_dict(checkpoint['state_dict'])
dataloader = ld.GazeFollow(opt)
images, xis, eye_coords, pred_coords, eyes, names, eyes2, gaze_final = next(
iter(dataloader.val_loader))
images, xis, eye_coords, pred_coords, gaze_final = images.cuda(), xis.cuda(
), eye_coords.cuda(), pred_coords.cuda(), gaze_final.cuda()
outputs = model.predict(images, xis, eye_coords)
untr = transforms.Compose(
[transforms.Normalize([0, 0, 0], [1 / (0.229), 1 / (0.224), 1 / (0.225)])])
untr2 = transforms.Compose(
sys.path.append('./sfd')
parser = opts.optionargparser()
opt = parser.parse_args()
torch.device('cpu')
# checkpoint = torch.load('./savedmodels/adamodels/gazenet_gazefollow_99epoch.pth.tar')
checkpoint = torch.load(
'./savedmodels/gazenet_gazefollow_highcrop_Hope_20epoch.pth.tar',
map_location=lambda storage, loc: storage)
print("Loading our pretrained gazenet model: ")
start_epoch = checkpoint['epoch']
#best_err = checkpoint['best_prec1']
model = gazenet.Net(opt).cpu()
model.load_state_dict(checkpoint['state_dict'])
imglist = glob.glob('imgs/test/*')
c = 0
for imname in imglist:
im = io.imread(imname)
faces, eye_coords, eyes = facedetect.getFaces(im)
normtransform = transforms.Compose(
[transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])])
untr = transforms.Compose([
transforms.Normalize([0, 0, 0],
[1 / (0.229), 1 / (0.224), 1 / (0.225)])