Spaces:
Runtime error
Runtime error
Added visualise.py for visualising the predictions
Browse files- visualise.py +98 -0
visualise.py
ADDED
@@ -0,0 +1,98 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Visualisation code for SMPL-X model. This code is useful if you already have predictions.
|
2 |
+
|
3 |
+
import os
|
4 |
+
import sys
|
5 |
+
import os.path as osp
|
6 |
+
import numpy as np
|
7 |
+
import smplx
|
8 |
+
from smplx.joint_names import JOINT_NAMES
|
9 |
+
import torch
|
10 |
+
try:
|
11 |
+
CUR_DIR = osp.dirname(os.path.abspath(__file__))
|
12 |
+
except NameError:
|
13 |
+
CUR_DIR = os.getcwd()
|
14 |
+
sys.path.insert(0, osp.join(CUR_DIR, '..', 'main'))
|
15 |
+
sys.path.insert(0, osp.join(CUR_DIR , '..', 'common'))
|
16 |
+
import matplotlib.pyplot as plt
|
17 |
+
from mpl_toolkits.mplot3d import Axes3D
|
18 |
+
|
19 |
+
JOINT_NAMES_DICT = {name: i for i, name in enumerate(JOINT_NAMES)}
|
20 |
+
|
21 |
+
# Load the SMPL-X model
|
22 |
+
model_path = 'common/utils/human_model_files' # Update with the path to your SMPL-X models
|
23 |
+
model = smplx.create(model_path, model_type='smplx', gender='neutral', ext='npz')
|
24 |
+
|
25 |
+
# Load the parameters from the .npz file
|
26 |
+
data = np.load('/home/sahand/Downloads/smplx/00047_9.npz')
|
27 |
+
|
28 |
+
betas = torch.tensor(data['betas'], dtype=torch.float32)
|
29 |
+
body_pose = torch.tensor(data['body_pose'], dtype=torch.float32)
|
30 |
+
global_orient = torch.tensor(data['global_orient'], dtype=torch.float32)
|
31 |
+
transl = torch.tensor(data['transl'], dtype=torch.float32)
|
32 |
+
expression = torch.tensor(data['expression'], dtype=torch.float32)
|
33 |
+
|
34 |
+
# Add missing dimensions to the tensors
|
35 |
+
if betas.ndim == 1:
|
36 |
+
betas = betas.unsqueeze(0)
|
37 |
+
if body_pose.ndim == 2:
|
38 |
+
body_pose = body_pose.unsqueeze(0)
|
39 |
+
if global_orient.ndim == 1:
|
40 |
+
global_orient = global_orient.unsqueeze(0)
|
41 |
+
if transl.ndim == 1:
|
42 |
+
transl = transl.unsqueeze(0)
|
43 |
+
if expression.ndim == 1:
|
44 |
+
expression = expression.unsqueeze(0)
|
45 |
+
|
46 |
+
# Reshape body_pose to include the batch dimension
|
47 |
+
body_pose = body_pose.view(1, -1, 3)
|
48 |
+
|
49 |
+
# Forward pass through the model
|
50 |
+
output = model(betas=betas, body_pose=body_pose, global_orient=global_orient, transl=transl, expression=expression)
|
51 |
+
|
52 |
+
# Extract joint positions
|
53 |
+
joints = output.joints.detach().cpu().numpy().squeeze()
|
54 |
+
print(joints.shape)
|
55 |
+
# Ankle joints (left and right)
|
56 |
+
left_knee = joints[4] # Index for left ankle in SMPL-X
|
57 |
+
right_knee = joints[5] # Index for right ankle in SMPL-X
|
58 |
+
left_ankle = joints[7] # Index for left ankle in SMPL-X
|
59 |
+
right_ankle = joints[8] # Index for right ankle in SMPL-X
|
60 |
+
|
61 |
+
bone_connections = [
|
62 |
+
(JOINT_NAMES_DICT["pelvis"], JOINT_NAMES_DICT["spine1"]), (JOINT_NAMES_DICT["spine1"], JOINT_NAMES_DICT["spine2"]), (JOINT_NAMES_DICT["spine2"], JOINT_NAMES_DICT["spine3"]), # Spine
|
63 |
+
(JOINT_NAMES_DICT["pelvis"], JOINT_NAMES_DICT["left_hip"]), (JOINT_NAMES_DICT["left_hip"], JOINT_NAMES_DICT["left_knee"]), (JOINT_NAMES_DICT["left_knee"], JOINT_NAMES_DICT["left_ankle"]), # Left leg
|
64 |
+
(JOINT_NAMES_DICT["pelvis"], JOINT_NAMES_DICT["right_hip"]), (JOINT_NAMES_DICT["right_hip"], JOINT_NAMES_DICT["right_knee"]), (JOINT_NAMES_DICT["right_knee"], JOINT_NAMES_DICT["right_ankle"]), # Right leg
|
65 |
+
(JOINT_NAMES_DICT["left_ankle"], JOINT_NAMES_DICT["left_heel"]),
|
66 |
+
(JOINT_NAMES_DICT["right_ankle"], JOINT_NAMES_DICT["right_heel"]),
|
67 |
+
(JOINT_NAMES_DICT["left_ankle"], JOINT_NAMES_DICT["left_foot"]),
|
68 |
+
(JOINT_NAMES_DICT["left_foot"], JOINT_NAMES_DICT["left_big_toe"]), (JOINT_NAMES_DICT["left_foot"], JOINT_NAMES_DICT["left_small_toe"]),
|
69 |
+
(JOINT_NAMES_DICT["right_ankle"], JOINT_NAMES_DICT["right_foot"]),
|
70 |
+
(JOINT_NAMES_DICT["right_foot"], JOINT_NAMES_DICT["right_big_toe"]), (JOINT_NAMES_DICT["right_foot"], JOINT_NAMES_DICT["right_small_toe"]),
|
71 |
+
# Add more bones if necessary
|
72 |
+
]
|
73 |
+
|
74 |
+
# Visualize the 3D skeleton
|
75 |
+
fig = plt.figure()
|
76 |
+
ax = fig.add_subplot(111, projection='3d')
|
77 |
+
|
78 |
+
# Plot all joints
|
79 |
+
ax.scatter(joints[:, 0], joints[:, 1], joints[:, 2], c='blue', marker='o')
|
80 |
+
# Highlight ankle joints
|
81 |
+
ax.scatter([left_knee[0]], [left_knee[1]], [left_knee[2]], c='red', marker='x', s=100, label='Left Knee')
|
82 |
+
ax.scatter([right_knee[0]], [right_knee[1]], [right_knee[2]], c='green', marker='x', s=100, label='Right Knee')
|
83 |
+
ax.scatter([left_ankle[0]], [left_ankle[1]], [left_ankle[2]], c='red', marker='o', s=100, label='Left Ankle')
|
84 |
+
ax.scatter([right_ankle[0]], [right_ankle[1]], [right_ankle[2]], c='green', marker='o', s=100, label='Right Ankle')
|
85 |
+
|
86 |
+
# Draw bones
|
87 |
+
for bone in bone_connections:
|
88 |
+
start, end = bone
|
89 |
+
ax.plot([joints[start, 0], joints[end, 0]],
|
90 |
+
[joints[start, 1], joints[end, 1]],
|
91 |
+
[joints[start, 2], joints[end, 2]], 'k-')
|
92 |
+
|
93 |
+
# Set labels
|
94 |
+
ax.set_xlabel('X')
|
95 |
+
ax.set_ylabel('Y')
|
96 |
+
ax.set_zlabel('Z')
|
97 |
+
ax.legend()
|
98 |
+
plt.show()
|