def run_model_single(gender, pose_params, shape_params, trans_params):
smpl_params = torch.Tensor(
torch.cat(
(pose_params, shape_params, trans_params)).type(torch.float32))
smpl_layer = SMPL_Layer(center_idx=0,
gender=gender,
model_root=settings.SMPL_MODEL_PATH)
verts, Jtr = smpl_layer(
torch.unsqueeze(smpl_params[:pose_size], 0),
th_betas=torch.unsqueeze(smpl_params[pose_size:pose_size + beta_size],
0),
th_trans=torch.unsqueeze(smpl_params[pose_size + beta_size:], 0))
return verts.numpy()[0], smpl_layer.th_faces.numpy()
def get_init_joint_2d(intrinsics, extrinsics, gender):
pose_params = torch.zeros(pose_size)
shape_params = torch.rand(beta_size)
trans_params = torch.tensor([0, 1, 0])
smpl_layer = SMPL_Layer(center_idx=0,
gender=gender,
model_root=settings.SMPL_MODEL_PATH)
smpl_params = torch.Tensor(
torch.cat(
(pose_params, shape_params, trans_params)).type(torch.float32))
verts, Jtr = smpl_layer(
torch.unsqueeze(smpl_params[:pose_size], 0),
th_betas=torch.unsqueeze(smpl_params[pose_size:pose_size + beta_size],
0),
th_trans=torch.unsqueeze(smpl_params[pose_size + beta_size:], 0))
cam_param = common_utils.get_camera_info(intrinsics, extrinsics)
pred_joint2d = common_utils.projection_torch(vertices=Jtr[0],
cam_param=cam_param,
width=1920,
height=1080)
return pred_joint2d.numpy()
import torch
from smplpytorch.pytorch.smpl_layer import SMPL_Layer
from display_utils import display_model
from pose_data import pose1, pose2, pose3, pose4
if __name__ == '__main__':
cuda = False
batch_size = 1
# Create the SMPL layer
smpl_layer = SMPL_Layer(center_idx=0,
gender='female',
model_root='smplpytorch/native/models')
userInp = input(
"Please enter the desired pose (Hint: interger number from [1-4] : ")
pose_params = None
if userInp == '1':
pose_params = pose1
elif userInp == '2':
pose_params = pose2
elif userInp == '3':
pose_params = pose3
elif userInp == '4':
pose_params = pose4
else:
print("Please enter values between 1 to 4")
# Random preset shape
shape_params = torch.tensor(
[[0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1]]) * 20
return args
# argument parsing
args = parse_args()
cfg.set_args(args.gpu_ids, args.stage)
cudnn.benchmark = True
# SMPL joint set
joint_num = 29 # original: 24. manually add nose, L/R eye, L/R ear
joints_name = ('Pelvis', 'L_Hip', 'R_Hip', 'Torso', 'L_Knee', 'R_Knee', 'Spine', 'L_Ankle', 'R_Ankle', 'Chest', 'L_Toe', 'R_Toe', 'Neck', 'L_Thorax', 'R_Thorax', 'Head', 'L_Shoulder', 'R_Shoulder', 'L_Elbow', 'R_Elbow', 'L_Wrist', 'R_Wrist', 'L_Hand', 'R_Hand', 'Nose', 'L_Eye', 'R_Eye', 'L_Ear', 'R_Ear')
flip_pairs = ( (1,2), (4,5), (7,8), (10,11), (13,14), (16,17), (18,19), (20,21), (22,23) , (25,26), (27,28) )
skeleton = ( (0,1), (1,4), (4,7), (7,10), (0,2), (2,5), (5,8), (8,11), (0,3), (3,6), (6,9), (9,14), (14,17), (17,19), (19, 21), (21,23), (9,13), (13,16), (16,18), (18,20), (20,22), (9,12), (12,24), (24,15), (24,25), (24,26), (25,27), (26,28) )
# SMPl mesh
vertex_num = 6890
smpl_layer = SMPL_Layer(gender='neutral', model_root=cfg.smpl_path + '/smplpytorch/native/models')
face = smpl_layer.th_faces.numpy()
joint_regressor = smpl_layer.th_J_regressor.numpy()
root_joint_idx = 0
# snapshot load
model_path = './snapshot_%d.pth.tar' % int(args.test_epoch)
assert osp.exists(model_path), 'Cannot find model at ' + model_path
print('Load checkpoint from {}'.format(model_path))
model = get_model(vertex_num, joint_num, 'test')
model = DataParallel(model).cuda()
ckpt = torch.load(model_path)
model.load_state_dict(ckpt['network'], strict=False)
model.eval()
def get_layer(self, gender='neutral'):
return SMPL_Layer(gender=gender, model_root=cfg.smpl_path + '/smplpytorch/native/models')
import torch
from smplpytorch.pytorch.smpl_layer import SMPL_Layer
from display_utils import display_model
if __name__ == '__main__':
cuda = False
batch_size = 1
# Create the SMPL layer
smpl_layer = SMPL_Layer(
center_idx=0,
gender='male',
model_root=
'/BS/bharat/work/installation/smplpytorch/smplpytorch/native/models')
# Generate random pose and shape parameters
pose_params = torch.rand(batch_size, 72) * 0.2
shape_params = torch.rand(batch_size, 10) * 0.03
offsets = torch.rand(batch_size, 6890, 3) * 0.05
# GPU mode
if cuda:
pose_params = pose_params.cuda()
shape_params = shape_params.cuda()
smpl_layer.cuda()
offsets.cuda()
# Forward from the SMPL layer
verts, Jtr = smpl_layer(pose_params,
th_betas=shape_params,
import torch
import cv2
import numpy as np
from smplpytorch.pytorch.smpl_layer import SMPL_Layer
from Test.smpl_pytorch.display_utils import display_model
import matplotlib.pyplot as plt
from pytorch3d.io import save_obj
if __name__ == '__main__':
cuda = False
batch_size = 1
# Create the SMPL layer
smpl_layer = SMPL_Layer(
center_idx=0,
gender='male',
model_root='/Users/anupamtripathi/PycharmProjects/3d_body_mesurement/Test/smpl_pytorch/smlppytorch/native/models')
# Generate random pose and shape parameters
pose_params = torch.ones(batch_size, 72) * 0
shape_params = torch.ones(batch_size, 10)
pose_params = torch.tensor([[0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., -1, 0., 0., 1, 0., 0., 0.,
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.]]).float()
# GPU mode
if cuda:
pose_params = pose_params.cuda()
shape_params = shape_params.cuda()
smpl_layer.cuda()
def get_layer(self, gender):
return SMPL_Layer(gender=gender,
model_root=cfg.smpl_dir +
'/smplpytorch/native/models')
def run_model(gt_dataset, gender, epochs=10):
# parameter setting
pose_params = torch.zeros(pose_size)
shape_params = torch.rand(beta_size)
trans_params = torch.tensor([0, 1, 0])
# trans_params = torch.zeros(3)
# trans_params = torch.zeros(3)
smpl_params = Variable(torch.cat(
(pose_params, shape_params, trans_params)).type(torch.float32),
requires_grad=True)
smpl_layer = SMPL_Layer(center_idx=0,
gender=gender,
model_root=settings.SMPL_MODEL_PATH)
# verts, Jtr = smpl_layer(pose_params, th_betas=shape_params, th_trans=trans_params)
# mesh = trimesh.Trimesh(vertice, faces)
# display_utils.save_obj(mesh.vertices, mesh.faces, 'test.obj')
datasets = []
for gt_joint2d, intrinsics, extrinsics in gt_dataset:
datasets.append({
'joint2d': torch.Tensor(gt_joint2d),
'intrinsics': torch.Tensor(intrinsics),
'extrinsics': torch.Tensor(extrinsics)
})
# hyper parameter
learning_rate = 1e-1
optimizer = optim.Adam([smpl_params], lr=learning_rate)
for epoch in tqdm(range(epochs)):
for dataset in datasets:
verts, Jtr = smpl_layer(
torch.unsqueeze(smpl_params[:pose_size], 0),
th_betas=torch.unsqueeze(
smpl_params[pose_size:pose_size + beta_size], 0),
th_trans=torch.unsqueeze(smpl_params[pose_size + beta_size:],
0))
gt_joint2d = dataset['joint2d']
intrinsics = dataset['intrinsics']
extrinsics = dataset['extrinsics']
cam_param = common_utils.get_camera_info(intrinsics, extrinsics)
pred_joint2d = common_utils.projection_torch(vertices=Jtr[0],
cam_param=cam_param,
width=1920,
height=1080)
loss = (pred_joint2d - gt_joint2d).pow(2).sum()
optimizer.zero_grad()
loss.backward()
optimizer.step()
pose_t = torch.unsqueeze(smpl_params[:pose_size],
0).reshape(-1, 3).detach().numpy()
rotation = []
for pose in pose_t:
r = R.from_rotvec(pose)
rotation.append(r.as_euler('xyz', degrees=True).tolist())
# th_pose_rotmat = th_posemap_axisang(pose_t).reshape(-1, 9)
# pose_t = th_pose_rotmat.detach().numpy()
result = {
'rotation':
rotation,
'shape_params':
smpl_params.detach().numpy()[pose_size:pose_size + beta_size].tolist(),
'trans_params':
smpl_params.detach().numpy()[pose_size + beta_size:].tolist(),
'joint_3d':
Jtr[0].detach().numpy().squeeze().tolist()
}
return verts, Jtr, result
import torch
from smplpytorch.pytorch.smpl_layer import SMPL_Layer
from display_utils import display_model
if __name__ == '__main__':
cuda = False
batch_size = 1
# Create the SMPL layer
smpl_layer = SMPL_Layer(center_idx=0,
gender='neutral',
model_root='models')
# Generate random pose and shape parameters
pose_params = torch.rand(batch_size, 72) * 0.2
shape_params = torch.rand(batch_size, 10) * 0.03
# GPU mode
if cuda:
pose_params = pose_params.cuda()
shape_params = shape_params.cuda()
smpl_layer.cuda()
# Forward from the SMPL layer
verts, Jtr = smpl_layer(pose_params, th_betas=shape_params)
# Draw output vertices and joints
display_model({
'verts': verts.cpu().detach(),
'joints': Jtr.cpu().detach()
