fish_tracking/analyze_tracking.py at main · drorki/fish_tracking · 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
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

import numpy as np
import matplotlib.pyplot as plt
import pickle


def convert(size,x,y,w,h):
    box = np.zeros(4)
    dw = 1./size[0]
    dh = 1./size[1]
    x = x/dw
    w = w/dw
    y = y/dh
    h = h/dh
    box[0] = x-(w/2.0)
    box[1] = x+(w/2.0)
    box[2] = y-(h/2.0)
    box[3] = y+(h/2.0)
    return (box)


def get_annotated_track(data_path, img_h, img_w):
    '''
    In yolo, a bounding box is represented by four values [x_center, y_center, width, height] . x_center and y_center
    are the normalized coordinates of the center of the bounding box. To make coordinates normalized, we take pixel
    values of x and y, which marks the center of the bounding box on the x- and y-axis
    '''
    n_frames = 816
    bbox = []
    xy = []
    for frame_no in range(n_frames):
        filename = data_path + r'\frame' + str(frame_no) + '.txt'
        txt_file = open(filename, "r")
        line = txt_file.read().splitlines()
        values = line[0].split()
        x = y = w = h = cls = None
        cls = line[0]
        x = float(values[1])
        y = float(values[2])
        w = float(values[3])
        h = float(values[4])
        bbox.append(convert((img_w, img_h), x, y, w, h))
        xy.append([x, y])
    return xy, bbox


def relative2pixels(xy_list, img_w, img_h):
    # convert relative coordinates in xy_list to pixels
    xy = np.array(xy_list)
    xy[:, 0] = xy[:, 0] * img_w  # x-coordinate in pixels
    xy[:, 1] = xy[:, 1] * img_h  # y-coordinate in pixels
    return xy

############## PARAMETERS ################################
img_h = 1296  # pixels
img_w = 2304  # pixels
dist_thd_pixels = 50
video_path = r'.\data\video'
#video_path = r'C:\Users\d_kip\work\lab_projects\Fish_Noise\Video\VIDEO_20230223_133606599'
annotations_path = r'\VIDEO_20230223_133606599_annotation\fish_bottom_right'
#annotations_path = video_path + r'\fish_first_from_right'
#annotations_path = video_path + r'\fish_first_left'
#annotations_path = video_path + r'\fish_second_from_left'
#annotations_path = video_path + r'\fish_second_from_right' # 'Original' annotation
#annotations_path = video_path + r'\fish_third_from_right'

###########################################################


#filename2load = 'tracks_20230228-173213.pkl'  # max_iou_dist = 0.5, max_age = 12, n_init=4, max_cosine_distance = 0.4 --> 147/816
#filename2load = 'tracks_20230228-181308.pkl'  # max_iou_dist = 0.2, max_age = 12, n_init=4, max_cosine_distance = 0.4 --> very few trackers
#filename2load = 'tracks_20230228-181822.pkl'  # max_iou_dist = 0.8, max_age = 12, n_init=4, max_cosine_distance = 0.4 --> 183/816
#filename2load = 'tracks_20230228-183435.pkl'  # max_iou_dist = 0.8, max_age = 12, n_init=4, max_cosine_distance = 0.8 --> 184/816
#filename2load = 'tracks_20230301-122144.pkl'  # ROI, 10000 epochs, max_cosine_distance = 0.4, max_age = 12, max_iou_dist = 0.5, n_init=4, nn_budget=None <-- 241 / 816
#filename2load = 'tracks_20230301-170515.pkl'  # ROI, 10000 epochs, max_cosine_distance = 0.4, max_age = 12, max_iou_dist = 0.5, n_init=3, nn_budget=None <--  282 / 816
#filename2load = 'tracks_20230301-171343.pkl'  # ROI, 10000 epochs, max_cosine_distance = 0.9, max_age = 12, max_iou_dist = 0.5, n_init=3, nn_budget=None <-- 293 / 816
#filename2load = 'tracks_20230301-172726.pkl'  # ROI, 10000 epochs, max_cosine_distance = 0.9, max_age = 12, max_iou_dist = 0.5, n_init=3, nn_budget=1 <-- 293 / 816
#filename2load = 'tracks_20230301-173333.pkl'  # ROI, 10000 epochs, max_cosine_distance = 0.9, max_age = 12, max_iou_dist = 0.8, n_init=3, nn_budget=None <-- 307 / 816
#filename2load = 'tracks_20230301-174037.pkl'  # ROI, 10000 epochs, max_cosine_distance = 0.9, max_age = 12, max_iou_dist = 0.8, n_init=6, nn_budget=None <-- 186 / 816
#filename2load = 'tracks_20230302-113337.pkl'  # ROI, 9000 epochs, max_cosine_distance = 0.9, max_age = 12, max_iou_dist = 0.8, n_init=3, nn_budget=None <--  483 / 816 (59%!)
#filename2load = 'tracks_20230302-114112.pkl'  # ROI, 9000 epochs, max_cosine_distance = 0.9, max_age = 100, max_iou_dist = 0.8, n_init=3, nn_budget=None <-- 576 / 816 looks like too-many irrelevant tracks
#filename2load = 'tracks_20230302-115356.pkl'  # ROI, 9000 epochs, max_cosine_distance = 0.9, max_age = 24, max_iou_dist = 0.8, n_init=3, nn_budget=None <--  514 / 816 (63%!)
#filename2load = 'tracks_20230302-120735.pkl'  # ROI, 9000 epochs, max_cosine_distance = 0.9, max_age = 24, max_iou_dist = 0.8, n_init=3, nn_budget=None, verify_detections <--  526 / 816 (64%!)
filename2load = 'tracks_20230302-122106.pkl'  # ROI, 15000 epochs, max_cosine_distance = 0.9, max_age = 24, max_iou_dist = 0.8, n_init=3, nn_budget=None, verify_detections <-- 515 / 816 (63%!)
# ------------------------------------------

if __name__ == '__main__':

    # Tracking data
    with open('performance_evaluation/' + filename2load, 'rb') as f:
        #history_dict, nn_budget, max_cosine_distance, max_age, max_iou_dist, frame_height, frame_width = pickle.load(f)
        history_dict, max_cosine_distance, nn_budget, max_age, max_iou_dist, n_init, roi, frame_height, frame_width, model_path = pickle.load(f)
    print('\nFile:', filename2load, ' ---------------')
    print('model_path =', model_path)
    print('max_cosine_distance =', max_cosine_distance)
    print('nn_budget =', nn_budget)
    print('max_age =', max_age)
    print('max_iou_dist =', max_iou_dist)
    print('n_init =', n_init)
    print('Number of valid trackers:', len(history_dict.keys()))
    print('Available trackers are:', history_dict.keys())

    # Annotation data
    print('annotations_path =', annotations_path)
    xy_list, bbox_list = get_annotated_track(video_path + annotations_path, img_h, img_w)
    xy = relative2pixels(xy_list, img_w, img_h)
    xy[:, 0] = xy[:, 0] - roi['top left xy'][0]  # compensate x for the top-left of teh ROI
    xy[:, 1] = xy[:, 1] - roi['top left xy'][1]  # compensate y for the top-left of teh ROI


    n_coverage_frames = 0
    good_tracks_count = 0  # Count number of tracks that follows the target at least 60% of their duration
    # Calculate coverage
    for key in history_dict.keys():
        xy_data = np.array(history_dict[key]['center'])
        frame = np.array(history_dict[key]['frame'])-2
        dist = np.abs(xy[frame, :] - xy_data)
        n_coverage_track = np.sum(np.bitwise_and(dist[:, 0] < dist_thd_pixels, dist[:, 1] < dist_thd_pixels))  # Count the number of frames during which the current track follows the fish
        good_tracks_count += n_coverage_track/xy_data.shape[0] > 0.6  # We consider a track as 'good' if at least 60% of it follows the fish
        print('Coverage of track', key, 'is', n_coverage_track, '/', xy_data.shape[0])
        n_coverage_frames += n_coverage_track
    print('dist_thd_pixels =', dist_thd_pixels, '--> n_coverage_frames =', n_coverage_frames, '/', xy.shape[0])
    print('Number of tracks which follows the fish at least 60% of the time:', good_tracks_count)
    # Plot x vs. frame number
    plt.figure().canvas.manager.set_window_title(annotations_path.split('\\')[-1])
    plt.plot(xy[:,0], '--')
    plt.fill_between(np.arange(len(xy)), xy[:,0]-dist_thd_pixels, xy[:,0]+dist_thd_pixels, color='b', alpha=.1)
    for key in history_dict.keys():
        xy_data = np.array(history_dict[key]['center'])
        frame = np.array(history_dict[key]['frame'])
        plt.plot(frame, xy_data[:,0], label=str(key))
    plt.xlabel('frame#')
    plt.ylabel('x')

    # Plot y vs. frame number
    plt.figure().canvas.manager.set_window_title(annotations_path.split('\\')[-1])
    plt.plot(xy[:,1], '--')
    plt.fill_between(np.arange(len(xy)), xy[:, 1] - dist_thd_pixels, xy[:, 1] + dist_thd_pixels, color='b', alpha=.1)
    for key in history_dict.keys():
        xy_data = np.array(history_dict[key]['center'])
        frame = np.array(history_dict[key]['frame'])
        plt.plot(frame, xy_data[:,1], label=str(key))
    plt.xlabel('frame#')
    plt.ylabel('y')

    plt.show()

'''
    # Plot (x,y)
    plt.figure().canvas.manager.set_window_title(annotations_path.split('\\')[-1])
    plt.plot(xy[:,0], xy[:,1], '.-')
    plt.xlabel('x')
    plt.ylabel('y')
    plt.show()
'''