tolerance_change=1e-7,
max_iter=1500,
line_search_fn='strong_wolfe')
Adam_optimizer = torch.optim.Adam(vars, lr=0.05)
print('silhouette fit')
plot_silhouette(flamelayer, renderer, target_silh)
# fitter.optimize_Adam(Adam_optimizer,1,1e-4)
fitter.optimize_LBFGS(all_flame_params_optimizer, 1, 1e-4)
plot_silhouette(flamelayer, renderer, target_silh)
if __name__ == '__main__':
parser.add_argument('--target_img_path',
type=str,
default='./data/bareteeth.000001.26_C.jpg',
help='Target image path')
parser.add_argument('--out_path',
type=str,
default='./Results',
help='Results folder path')
parser.add_argument('--texture_mapping',
type=str,
default='./data/texture_data.npy',
help='Texture data')
config = get_config()
config.batch_size = 1
config.flame_model_path = './model/male_model.pkl'
frames = video_to_images(inp, max_images)
output_file_paths = output_file_paths[:len(frames)]
for i in range(len(output_file_paths)):
if (use_greenscreen):
frames[i] = greenscreen_bg_to_black(frames[i])
cv2.imwrite(output_file_paths[i], frames[i])
else:
raise FileNotFoundError(errno.ENOENT, os.strerror(errno.ENOENT),
config.input)
return output_file_paths
if __name__ == '__main__':
parser.add_argument(
'--input',
help='Path of the input folder for images, or path of video file')
parser.add_argument('--output_folder', help='Output folder path')
parser.add_argument('--image_viewpoint_ending',
default='26_C.jpg',
help='Ending of the file from the given angle')
parser.add_argument('--texture_mapping',
type=str,
default='./data/texture_data.npy',
help='Texture data')
parser.add_argument('--load_shape_path',
type=str,
default='',
help='Load shape from a given path')
parser.add_argument('--max_images',
type=int,
val_metrics = model.compute_metrics(embeddings, data, 'val')
else:
n_warmup = 50
n_sample = 50
model.eval() # set evaluation mode
print("=== Running Warmup Passes")
for i in range(0,n_warmup):
embeddings = model.encode(data['features'], data['adj_train_norm'])
val_metrics = model.compute_metrics(embeddings, data, 'val')
print("=== Collecting Runtime over ", str(n_sample), " Passes")
tic = time.perf_counter()
for i in range(0,n_sample):
embeddings = model.encode(data['features'], data['adj_train_norm'])
val_metrics = model.compute_metrics(embeddings, data, 'val')
toc = time.perf_counter()
avg_runtime = float(toc - tic)/n_sample
print("average runtime = ", avg_runtime)
# write runtime to file
f = open(args.time_file, "w")
f.write(str(avg_runtime)+"\n")
f.close()
if __name__ == '__main__':
parser.add_argument('--time_file', type=str, default='', help='timing output file')
args = parser.parse_args()
profiler.start()
test(args)
profiler.stop()
net, args, epoch,
os.path.join(args.checkpoint_dir, 'epoch_%d.pth' % epoch))
def adjust_learning_rate(optimizer, lr):
new_lr = lr / 10
for param_group in optimizer.param_groups:
param_group['lr'] = new_lr
print('Adjusting learning rate, new lr is %f' % new_lr)
return new_lr
if __name__ == '__main__':
from config import parser
parser.add_argument('--resume',
default=None,
type=str,
help='Resume from checkpoint')
parser.add_argument('--gpu',
default='0',
type=str,
help='Which GPU to run on')
parser.add_argument('--save_folder',
default='weights/',
help='Location to save checkpoint models')
parser.add_argument('--epochs',
default=100,
type=int,
help='Maximum training epochs')
parser.add_argument('--train_data',
required=True,
help='Path to training data')