-
Notifications
You must be signed in to change notification settings - Fork 0
/
eval_d211227_coco_pymaf.py
285 lines (239 loc) · 12.8 KB
/
eval_d211227_coco_pymaf.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
#!/usr/bin/env python
# encoding: utf-8
'''
@project : pymaf_reimp
@file : eval_d211227_coco_pymaf.py
@author : Levon
@contact : levondang@163.com
@ide : PyCharm
@time : 2021-12-29 15:08
'''
"""
This script can be used to evaluate a trained model on 3D pose/shape and masks/part segmentation. You first need to download the datasets and preprocess them.
Example usage:
```
python3 eval_coco.py --checkpoint=data/pretrained_model/PyMAF_model_checkpoint.pt
```
Running the above command will compute the 2D keypoint detection error. The ```--dataset``` option can take different values based on the type of evaluation you want to perform:
1. COCO ```--dataset=coco```
"""
import os.path as osp
import cv2
import torch
import argparse
import numpy as np
from tqdm import tqdm
from torch.utils.data import DataLoader
from pprint import pprint
import platform
from d211227_pymaf_reimp.datas import COCODataset
from d211227_pymaf_reimp.nets import SMPL, PyMAF
from d211227_pymaf_reimp.utils.geometry import perspective_projection
from d211227_pymaf_reimp.utils.transforms import transform_preds
from d211227_pymaf_reimp.utils.uv_vis import vis_smpl_iuv
from d211227_pymaf_reimp.cfgs import ConfigPymaf
def run_evaluation(model, dataset, result_file,
batch_size=32, img_res=224,
num_workers=32, shuffle=False, output_dir=None):
"""Run evaluation on the datasets and metrics we report in the paper. """
model.eval()
device = cfg.device
# Transfer model to the GPU
model.to(device)
# Load SMPL model
smpl_neutral = SMPL(cfg.JOINT_MAP, cfg.JOINT_NAMES, cfg.J24_TO_J19, cfg.JOINT_REGRESSOR_TRAIN_EXTRA, cfg.SMPL_MODEL_DIR, create_transl=False).to(device)
save_results = result_file is not None
# Disable shuffling if you want to save the results
if save_results:
shuffle = False
# Create dataloader for the dataset
data_loader = DataLoader(dataset, batch_size=batch_size, shuffle=shuffle, num_workers=num_workers)
# Store SMPL parameters
smpl_pose = np.zeros((len(dataset), 72))
smpl_betas = np.zeros((len(dataset), 10))
smpl_camera = np.zeros((len(dataset), 3))
pred_joints = np.zeros((len(dataset), 17, 3))
num_joints = 17
num_samples = len(dataset)
print('dataset length: {}'.format(num_samples))
all_preds = np.zeros(
(num_samples, num_joints, 3),
dtype=np.float32
)
all_boxes = np.zeros((num_samples, 6))
image_path = []
filenames = []
imgnums = []
idx = 0
with torch.no_grad():
for _, batch in enumerate(tqdm(data_loader, desc='Eval', total=len(data_loader))):
if len(args.vis_imname) > 0:
imgnames = [i_n.split('/')[-1] for i_n in batch['imgname']]
name_hit = False
for i_n in imgnames:
if args.vis_imname in i_n:
name_hit = True
print('vis: ' + i_n)
if not name_hit:
continue
images = batch['img'].to(device)
scale = batch['scale'].numpy()
center = batch['center'].numpy()
num_images = images.size(0)
gt_keypoints_2d = batch['keypoints'] # 2D keypoints
# De-normalize 2D keypoints from [-1,1] to pixel space
gt_keypoints_2d_orig = gt_keypoints_2d.clone()
gt_keypoints_2d_orig[:, :, :-1] = 0.5 * img_res * (gt_keypoints_2d_orig[:, :, :-1] + 1)
preds_dict, _ = model(images)
pred_rotmat = preds_dict['smpl_out'][-1]['rotmat'].contiguous().view(-1, 24, 3, 3)
pred_betas = preds_dict['smpl_out'][-1]['theta'][:, 3:13].contiguous()
pred_camera = preds_dict['smpl_out'][-1]['theta'][:, :3].contiguous()
pred_output = smpl_neutral(betas=pred_betas, body_pose=pred_rotmat[:, 1:],
global_orient=pred_rotmat[:, 0].unsqueeze(1), pose2rot=False)
# pred_vertices = pred_output.vertices
pred_J24 = pred_output.joints[:, -24:]
pred_JCOCO = pred_J24[:, cfg.J24_TO_JCOCO]
# Convert Weak Perspective Camera [s, tx, ty] to camera translation [tx, ty, tz] in 3D given the bounding box size
# This camera translation can be used in a full perspective projection
pred_cam_t = torch.stack([pred_camera[:, 1],
pred_camera[:, 2],
2 * cfg.FOCAL_LENGTH / (img_res * pred_camera[:, 0] + 1e-9)], dim=-1)
camera_center = torch.zeros(len(pred_JCOCO), 2, device=pred_camera.device)
pred_keypoints_2d = perspective_projection(pred_JCOCO,
rotation=torch.eye(3, device=pred_camera.device).unsqueeze(
0).expand(len(pred_JCOCO), -1, -1),
translation=pred_cam_t,
focal_length=cfg.FOCAL_LENGTH,
camera_center=camera_center)
coords = pred_keypoints_2d + (img_res / 2.)
coords = coords.cpu().numpy()
gt_keypoints_coco = gt_keypoints_2d_orig[:, -24:][:, cfg.J24_TO_JCOCO]
vert_errors_batch = []
for i, (gt2d, pred2d) in enumerate(zip(gt_keypoints_coco.cpu().numpy(), coords.copy())):
vert_error = np.sqrt(np.sum((gt2d[:, :2] - pred2d[:, :2]) ** 2, axis=1))
vert_error *= gt2d[:, 2]
vert_mean_error = np.sum(vert_error) / np.sum(gt2d[:, 2] > 0)
vert_errors_batch.append(10 * vert_mean_error)
preds = coords.copy()
scale_ = np.array([scale, scale]).transpose()
# Transform back
for i in range(coords.shape[0]):
preds[i] = transform_preds(
coords[i], center[i], scale_[i], [img_res, img_res]
)
all_preds[idx:idx + num_images, :, 0:2] = preds[:, :, 0:2]
all_preds[idx:idx + num_images, :, 2:3] = 1.
all_boxes[idx:idx + num_images, 5] = 1.
image_path.extend(batch['imgname'])
idx += num_images
if len(args.vis_imname) > 0:
exit()
osv = platform.system().lower()
if osv == "windows":
ckp_name = args.checkpoint.split('\\')
elif osv == "linux":
ckp_name = args.checkpoint.split('/')
ckp_name = ckp_name[-1].split('_')[0]
name_values, perf_indicator = dataset.evaluate(
all_preds, output_dir, all_boxes, image_path, ckp_name, filenames, imgnums
) # preds, output_dir, all_boxes, img_path, ckp_name
model_name = args.regressor
if isinstance(name_values, list):
for name_value in name_values:
_print_name_value(name_value, model_name)
else:
_print_name_value(name_values, model_name)
# Save reconstructions to a file for further processing
if save_results:
np.savez(result_file, pred_joints=pred_joints, pose=smpl_pose, betas=smpl_betas, camera=smpl_camera)
# markdown format output
def _print_name_value(name_value, full_arch_name):
names = name_value.keys()
values = name_value.values()
num_values = len(name_value)
print(
'| Arch ' +
' '.join(['| {}'.format(name) for name in names]) +
' |'
)
print('|---' * (num_values + 1) + '|')
if len(full_arch_name) > 15:
full_arch_name = full_arch_name[:8] + '...'
print(
'| ' + full_arch_name + ' ' +
' '.join(['| {:.3f}'.format(value) for value in values]) +
' |'
)
if __name__ == '__main__':
# Define command-line arguments
parser = argparse.ArgumentParser()
parser.add_argument('--dataset', default='coco', help='Choose evaluation dataset')
# parser.add_argument('--batch_size', default=32, type=int, help='Batch size for testing')
parser.add_argument('--batch_size', default=8, type=int, help='Batch size for testing')
parser.add_argument('--regressor', type=str, choices=['hmr', 'pymaf_net'], default='pymaf_net',
help='Name of the SMPL regressor.')
parser.add_argument('--result_file', default=None, help='If set, save detections to a .npz file')
parser.add_argument('--vis_demo', default=False, action='store_true', help='result visualization')
parser.add_argument('--ratio', default=1, type=int, help='image size ration for visualization')
parser.add_argument('--vis_imname', type=str, default='', help='image name used for visualization.')
parser.add_argument('--is_debug', default=False, type=bool, help='')
parser.add_argument('--checkpoint', default=r"H:\datas\three_dimension_reconstruction\spin_pymaf_data\pretrained_model\PyMAF_model_checkpoint.pt", help='Path to network checkpoint')
args = parser.parse_args()
cfg = ConfigPymaf()
print("\n================== Arguments =================")
pprint(vars(args), indent=4)
print("==========================================\n")
print("\n================== Configs =================")
pprint(vars(cfg), indent=4)
print("==========================================\n")
# 模型
# PyMAF model
model = PyMAF(cfg.pymaf_model['BACKBONE'], cfg.res_model['DECONV_WITH_BIAS'],
cfg.res_model['NUM_DECONV_LAYERS'], cfg.res_model['NUM_DECONV_FILTERS'],
cfg.res_model['NUM_DECONV_KERNELS'], cfg.pymaf_model['MLP_DIM'],
cfg.pymaf_model['N_ITER'], cfg.pymaf_model['AUX_SUPV_ON'], cfg.BN_MOMENTUM,
cfg.SMPL_MODEL_DIR, cfg.H36M_TO_J14, cfg.LOSS['POINT_REGRESSION_WEIGHTS'],
JOINT_MAP=cfg.JOINT_MAP, JOINT_NAMES=cfg.JOINT_NAMES, J24_TO_J19=cfg.J24_TO_J19,
JOINT_REGRESSOR_TRAIN_EXTRA=cfg.JOINT_REGRESSOR_TRAIN_EXTRA,
device=cfg.device, SMPL_MEAN_PARAMS_PATH=cfg.SMPL_MEAN_PARAMS_PATH, pretrained=True,
data_dir=cfg.preprocessed_data_dir)
if args.checkpoint is not None:
checkpoint = torch.load(args.checkpoint)
model.load_state_dict(checkpoint['model'], strict=True)
print(f"loaded from {args.checkpoint}!")
# Setup evaluation dataset
dataset = COCODataset(eval_pve=cfg.eval_pve, noise_factor=cfg.noise_factor, rot_factor=cfg.rot_factor, scale_factor=cfg.scale_factor,
ds=args.dataset, subset="val2014", ignore_3d=False, use_augmentation=True, is_train=False, is_debug=args.is_debug, DATASET_FOLDERS=cfg.ORIGIN_IMGS_DATASET_FOLDERS,
DATASET_FILES=cfg.PREPROCESSED_DATASET_FILES, JOINT_MAP=cfg.JOINT_MAP, JOINT_NAMES=cfg.JOINT_NAMES, J24_TO_J19=cfg.J24_TO_J19,
JOINT_REGRESSOR_TRAIN_EXTRA=cfg.JOINT_REGRESSOR_TRAIN_EXTRA, SMPL_MODEL_DIR=cfg.SMPL_MODEL_DIR, IMG_NORM_MEAN=cfg.IMG_NORM_MEAN,
IMG_NORM_STD=cfg.IMG_NORM_STD, TRAIN_BATCH_SIZE=cfg.TRAIN_BATCHSIZE, IMG_RES=cfg.IMG_RES, SMPL_JOINTS_FLIP_PERM=cfg.SMPL_JOINTS_FLIP_PERM, SMPL_POSE_FLIP_PERM=cfg.SMPL_POSE_FLIP_PERM)
# Run evaluation
args.result_file = None
run_evaluation(model, dataset, args.result_file, batch_size=args.batch_size, shuffle=False, num_workers=cfg.num_works, output_dir=cfg.output_dir)
print('{}: {}, {}'.format(args.regressor, args.checkpoint, args.dataset))
'''
=> writing results json to ./results/keypoints_val2014_results_PyMAF.json
Loading and preparing results...
DONE (t=3.31s)
creating index...
index created!
Running per image evaluation...
Evaluate annotation type *keypoints*
DONE (t=26.13s).
Accumulating evaluation results...
DONE (t=0.65s).
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets= 20 ] = 0.246
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets= 20 ] = 0.489
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets= 20 ] = 0.227
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets= 20 ] = 0.260
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets= 20 ] = 0.242
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 20 ] = 0.417
Average Recall (AR) @[ IoU=0.50 | area= all | maxDets= 20 ] = 0.672
Average Recall (AR) @[ IoU=0.75 | area= all | maxDets= 20 ] = 0.445
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets= 20 ] = 0.395
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets= 20 ] = 0.447
| Arch | AP | Ap .5 | AP .75 | AP (M) | AP (L) | AR | AR .5 | AR .75 | AR (M) | AR (L) |
|---|---|---|---|---|---|---|---|---|---|---|
| pymaf_net | 0.246 | 0.489 | 0.227 | 0.260 | 0.242 | 0.417 | 0.672 | 0.445 | 0.395 | 0.447 |
pymaf_net: H:\datas\three_dimension_reconstruction\spin_pymaf_data\pretrained_model\PyMAF_model_checkpoint.pt, coco
'''