BionicVision/saliency_priority.py at main · Azalea-Kim/BionicVision · GitHub

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
"""
Author: Yanxiu Jin
Date: 2025-03-17
Description: Create a circle mask around high saliency region and highlight pixels in the mask
Initial try for saliency priority, not final
"""
import matplotlib.pyplot as plt

import cv2
import numpy as np

def create_circle_mask_from_sal(sal_fill):
    if len(sal_fill.shape) == 3 and sal_fill.shape[2] == 3:
        sal_fill = cv2.cvtColor(sal_fill, cv2.COLOR_BGR2GRAY)
    sal_fill = sal_fill.astype(np.uint8)

    contours, _ = cv2.findContours(sal_fill, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    if not contours:
        return np.zeros_like(sal_fill, dtype=np.uint8)
    largest_contour = max(contours, key=cv2.contourArea)

    if len(largest_contour) < 5:
        return np.zeros_like(sal_fill, dtype=np.uint8)

    (cx, cy), radius = cv2.minEnclosingCircle(largest_contour)

    circle_mask = np.zeros_like(sal_fill, dtype=np.uint8)
    cv2.circle(circle_mask, (int(cx), int(cy)), int(radius), 255, thickness=-1)

    return circle_mask

seg_img = cv2.imread("D:\\2021-han-scene-simplification-master\\2021-han-scene-simplification-master\\segmentation_output\\kitchen_20fps_prioritize_noscene_final_png\\frame_163_seg.png",
                     cv2.IMREAD_GRAYSCALE)

seg = np.uint8(seg_img)

sal_img = cv2.imread("D:\\2021-han-scene-simplification-master\\2021-han-scene-simplification-master\\gaze_estimations\\kitchen_20fps\\gray\\gray_frame_163.jpg"
                     , cv2.IMREAD_GRAYSCALE)
sal = np.uint8(sal_img)


# Threshold the saliency map
sal_fil = sal.copy()
threshold = np.max(sal_fil) * .40  # .90
# threshold = np.percentile(sal, 90)
# print(threshold)
sal_fil[sal_fil <= threshold] = 0
sal_fil[sal_fil > 0] = 255


sal_fil1 = sal_fil
# Create the elliptical mask
sal_fil = create_circle_mask_from_sal(sal_fil)

# Visualize
plt.figure(figsize=(12, 5))
plt.subplot(1, 2, 1)
plt.imshow(sal_fil1, cmap="gray")
plt.title("Original Binary Mask (sal_fill)")
plt.axis("off")

plt.subplot(1, 2, 2)
plt.imshow(sal_fil, cmap="gray")
plt.title("Fitted Ellipse Mask")
plt.axis("off")

plt.show()

seg_sal = np.maximum(seg,sal_fil)
seg_sal = np.where(seg == 0, seg, seg_sal)


intersection = np.where((seg > 0) & (sal_fil > 0), 1, 0).astype(np.uint8)
# If seg is already uint8 in [0..255], convert to 3-channel
seg_rgb = cv2.cvtColor(seg, cv2.COLOR_GRAY2BGR)  # shape (H, W, 3)
# Mark intersection pixels in red (BGR: (0,0,255))


seg_rgb[intersection == 1] = (0, 0, 255)
plt.figure(figsize=(8,6))
plt.imshow(cv2.cvtColor(seg_rgb, cv2.COLOR_BGR2RGB))  # Convert BGR -> RGB for Matplotlib
plt.title("Intersection in Red")
plt.axis("off")
plt.show()


plt.imshow(seg_sal, cmap="gray", vmin=0, vmax=255)
plt.axis("off")
plt.title('seg_sal')
plt.show()