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
|
import numpy as np
import cv2
from PIL import Image
import scripts.r_masking.core as core
from reactor_utils import tensor_to_pil
try:
from ultralytics import YOLO
except Exception as e:
print(e)
def load_yolo(model_path: str):
try:
return YOLO(model_path)
except ModuleNotFoundError:
# https://github.com/ultralytics/ultralytics/issues/3856
YOLO("yolov8n.pt")
return YOLO(model_path)
def inference_bbox(
model,
image: Image.Image,
confidence: float = 0.3,
device: str = "",
):
pred = model(image, conf=confidence, device=device)
bboxes = pred[0].boxes.xyxy.cpu().numpy()
cv2_image = np.array(image)
if len(cv2_image.shape) == 3:
cv2_image = cv2_image[:, :, ::-1].copy() # Convert RGB to BGR for cv2 processing
else:
# Handle the grayscale image here
# For example, you might want to convert it to a 3-channel grayscale image for consistency:
cv2_image = cv2.cvtColor(cv2_image, cv2.COLOR_GRAY2BGR)
cv2_gray = cv2.cvtColor(cv2_image, cv2.COLOR_BGR2GRAY)
segms = []
for x0, y0, x1, y1 in bboxes:
cv2_mask = np.zeros(cv2_gray.shape, np.uint8)
cv2.rectangle(cv2_mask, (int(x0), int(y0)), (int(x1), int(y1)), 255, -1)
cv2_mask_bool = cv2_mask.astype(bool)
segms.append(cv2_mask_bool)
n, m = bboxes.shape
if n == 0:
return [[], [], [], []]
results = [[], [], [], []]
for i in range(len(bboxes)):
results[0].append(pred[0].names[int(pred[0].boxes[i].cls.item())])
results[1].append(bboxes[i])
results[2].append(segms[i])
results[3].append(pred[0].boxes[i].conf.cpu().numpy())
return results
class UltraBBoxDetector:
bbox_model = None
def __init__(self, bbox_model):
self.bbox_model = bbox_model
def detect(self, image, threshold, dilation, crop_factor, drop_size=1, detailer_hook=None):
drop_size = max(drop_size, 1)
detected_results = inference_bbox(self.bbox_model, tensor_to_pil(image), threshold)
segmasks = core.create_segmasks(detected_results)
if dilation > 0:
segmasks = core.dilate_masks(segmasks, dilation)
items = []
h = image.shape[1]
w = image.shape[2]
for x, label in zip(segmasks, detected_results[0]):
item_bbox = x[0]
item_mask = x[1]
y1, x1, y2, x2 = item_bbox
if x2 - x1 > drop_size and y2 - y1 > drop_size: # minimum dimension must be (2,2) to avoid squeeze issue
crop_region = core.make_crop_region(w, h, item_bbox, crop_factor)
if detailer_hook is not None:
crop_region = detailer_hook.post_crop_region(w, h, item_bbox, crop_region)
cropped_image = core.crop_image(image, crop_region)
cropped_mask = core.crop_ndarray2(item_mask, crop_region)
confidence = x[2]
# bbox_size = (item_bbox[2]-item_bbox[0],item_bbox[3]-item_bbox[1]) # (w,h)
item = core.SEG(cropped_image, cropped_mask, confidence, crop_region, item_bbox, label, None)
items.append(item)
shape = image.shape[1], image.shape[2]
segs = shape, items
if detailer_hook is not None and hasattr(detailer_hook, "post_detection"):
segs = detailer_hook.post_detection(segs)
return segs
def detect_combined(self, image, threshold, dilation):
detected_results = inference_bbox(self.bbox_model, core.tensor2pil(image), threshold)
segmasks = core.create_segmasks(detected_results)
if dilation > 0:
segmasks = core.dilate_masks(segmasks, dilation)
return core.combine_masks(segmasks)
def setAux(self, x):
pass
|
