forked from triple-Mu/YOLOv8-TensorRT
-
Notifications
You must be signed in to change notification settings - Fork 0
/
infer-seg.py
106 lines (91 loc) · 3.78 KB
/
infer-seg.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
from models import TRTModule # isort:skip
import argparse
from pathlib import Path
import cv2
import numpy as np
import torch
from config import ALPHA, CLASSES, COLORS, MASK_COLORS
from models.torch_utils import seg_postprocess
from models.utils import blob, letterbox, path_to_list
def main(args: argparse.Namespace) -> None:
device = torch.device(args.device)
Engine = TRTModule(args.engine, device)
H, W = Engine.inp_info[0].shape[-2:]
# set desired output names order
Engine.set_desired(['outputs', 'proto'])
images = path_to_list(args.imgs)
save_path = Path(args.out_dir)
if not args.show and not save_path.exists():
save_path.mkdir(parents=True, exist_ok=True)
for image in images:
save_image = save_path / image.name
bgr = cv2.imread(str(image))
draw = bgr.copy()
bgr, ratio, dwdh = letterbox(bgr, (W, H))
dw, dh = int(dwdh[0]), int(dwdh[1])
rgb = cv2.cvtColor(bgr, cv2.COLOR_BGR2RGB)
tensor, seg_img = blob(rgb, return_seg=True)
dwdh = torch.asarray(dwdh * 2, dtype=torch.float32, device=device)
tensor = torch.asarray(tensor, device=device)
# inference
data = Engine(tensor)
seg_img = torch.asarray(seg_img[dh:H - dh, dw:W - dw, [2, 1, 0]],
device=device)
bboxes, scores, labels, masks = seg_postprocess(
data, bgr.shape[:2], args.conf_thres, args.iou_thres)
masks = masks[:, dh:H - dh, dw:W - dw, :]
indices = (labels % len(MASK_COLORS)).long()
mask_colors = torch.asarray(MASK_COLORS, device=device)[indices]
mask_colors = mask_colors.view(-1, 1, 1, 3) * ALPHA
mask_colors = masks @ mask_colors
inv_alph_masks = (1 - masks * 0.5).cumprod(0)
mcs = (mask_colors * inv_alph_masks).sum(0) * 2
seg_img = (seg_img * inv_alph_masks[-1] + mcs) * 255
draw = cv2.resize(seg_img.cpu().numpy().astype(np.uint8),
draw.shape[:2][::-1])
bboxes -= dwdh
bboxes /= ratio
for (bbox, score, label) in zip(bboxes, scores, labels):
bbox = bbox.round().int().tolist()
cls_id = int(label)
cls = CLASSES[cls_id]
color = COLORS[cls]
cv2.rectangle(draw, bbox[:2], bbox[2:], color, 2)
cv2.putText(draw,
f'{cls}:{score:.3f}', (bbox[0], bbox[1] - 2),
cv2.FONT_HERSHEY_SIMPLEX,
0.75, [225, 255, 255],
thickness=2)
if args.show:
cv2.imshow('result', draw)
cv2.waitKey(0)
else:
cv2.imwrite(str(save_image), draw)
def parse_args() -> argparse.Namespace:
parser = argparse.ArgumentParser()
parser.add_argument('--engine', type=str, help='Engine file')
parser.add_argument('--imgs', type=str, help='Images file')
parser.add_argument('--show',
action='store_true',
help='Show the detection results')
parser.add_argument('--out-dir',
type=str,
default='./output',
help='Path to output file')
parser.add_argument('--conf-thres',
type=float,
default=0.25,
help='Confidence threshold')
parser.add_argument('--iou-thres',
type=float,
default=0.65,
help='Confidence threshold')
parser.add_argument('--device',
type=str,
default='cuda:0',
help='TensorRT infer device')
args = parser.parse_args()
return args
if __name__ == '__main__':
args = parse_args()
main(args)