detections.append(landmark32)
detections.append(cls16)
detections.append(bbox16)
detections.append(landmark16)
detections.append(cls8)
detections.append(bbox8)
detections.append(landmark8)
return detections
net = RetinaFace_MobileNet()
net.cuda()
from torchsummary import summary
summary(net, input_size=(3, 640, 640))
import pdb
pdb.set_trace()
img_dim = 640
input_size = (1, 3, img_dim, img_dim)
img = torch.FloatTensor(input_size[0], input_size[1], input_size[2],
input_size[3])
net = add_flops_counting_methods(net)
net.start_flops_count()
feat = net(img)
faceboxes_flops = net.compute_average_flops_cost()
print('Net Flops: {}'.format(flops_to_string(faceboxes_flops)))
print('Net Params: ' + get_model_parameters_number(net))
img_save_name = "1_Handshaking_Handshaking_1_579_result_epoch.jpg"
use_cuda = True
img_origin = np.float32(cv2.imread(img_name, cv2.IMREAD_COLOR))
net = CenterFace(phase='test', cfg=net_cfg)
# flops and params estimation
img_dim = 640
input_size = (1, 3, img_dim, img_dim)
img = torch.FloatTensor(input_size[0], input_size[1], input_size[2], input_size[3])
net = add_flops_counting_methods(net)
net.start_flops_count()
feat = net(img)
flops = net.compute_average_flops_cost()
print('Net Flops: {}'.format(flops_to_string(flops)))
print('Net Params: ' + get_model_parameters_number(net))
# load model
net = load_model(net, trained_model)
net.eval()
print('Finished loading model!')
# preprocess image
resize = 1600 / img_origin.shape[0]
img = cv2.resize(img_origin, None, None, fx=resize, fy=resize, interpolation=cv2.INTER_LINEAR)
im_height, im_width, _ = img.shape
scale = torch.Tensor([img.shape[1], img.shape[0], img.shape[1], img.shape[0]])
img -= (104, 117, 123)
img = img.transpose(2, 0, 1)
img = torch.from_numpy(img).unsqueeze(0)
if use_cuda:
img = img.cuda()