SWAG3/swag_tracker.py at master · banda13/SWAG3 · 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
import time
import numpy as np
from scipy.spatial import distance as dist, distance
from collections import OrderedDict
from swag_model import SwagModel
from swag_utils import calculate_centroids


class SwagTracker:

    def __init__(self, logger):
        self.next_object_ID = 0
        self.next_appear_ID = 0
        self.logger = logger
        self.avg_execution_sec = 0

        # TODO customize these params based on the current clip
        self.min_visibility = 10 # this depend on the fps
        self.max_disappeared = 10 # this have to depend on the place
        self.min_registration_distance = 100 # TODO optimize, 100 it to much I think

        self.all_tracked_objects = []
        self.current_objects = OrderedDict()
        self.disappeared_objects = OrderedDict()

        self.counter = 0
        self.current_objects_counter = 0

    def register(self, centroid, box, time_data):
        for (_, obj) in self.current_objects.items():
            d = distance.euclidean(centroid, obj.centroids)
            if d < self.min_registration_distance:
                # print("Avoiding double registration with distance: {}".format(d))
                return
        self.current_objects[self.next_object_ID] = SwagModel(self.next_object_ID, centroid, box, time_data)
        self.disappeared_objects[self.next_object_ID] = 0
        self.next_object_ID += 1

    def deregister(self, objectID, time_data):
        self.current_objects[objectID].last_appear = time_data
        if self.current_objects[objectID].appearance_counter > 3:
            self.current_objects[objectID].count_me = True
        self.all_tracked_objects.append(self.current_objects[objectID])

        del self.current_objects[objectID]
        del self.disappeared_objects[objectID]
        self.current_objects_counter -= 1

    def update(self, yolo_datas, time_data, frame_id):
        start_t = time.time()
        if len(yolo_datas) == 0:
            for objectID in list(self.disappeared_objects.keys()):
                self.disappeared_objects[objectID] += 1
                self.current_objects[objectID].predict_new_centroid(frame_id)

                if self.disappeared_objects[objectID] > self.max_disappeared:
                    self.deregister(objectID, time_data)

            self.avg_execution_sec = (self.avg_execution_sec + (time.time() - start_t)) / 2
            return self.current_objects

        input_centroids = np.zeros((len(yolo_datas), 2), dtype="int")
        input_labels = []
        input_confidences = []
        input_boxes = []

        for (i, d) in enumerate(yolo_datas):
            (startX, startY, endX, endY) = d[0]
            input_centroids[i] = calculate_centroids(startX, endX, startY, endY)
            input_labels.append(d[1])
            input_confidences.append(d[2])
            input_boxes.append(d[0])

        if len(self.current_objects) == 0:
            for i in range(0, len(input_centroids)):
                self.register(input_centroids[i], input_boxes[i], time_data)
        else:
            object_ids = list(self.current_objects.keys())
            object_centroids = [i.centroids for i in self.current_objects.values()]

            D = dist.cdist(np.array(object_centroids), input_centroids)
            rows = D.min(axis=1).argsort()
            cols = D.argmin(axis=1)[rows]
            used_rows = set()
            used_cols = set()
            for (row, col) in zip(rows, cols):
                if row in used_rows or col in used_cols:
                    continue

                objectID = object_ids[row]
                appearance_counter = self.current_objects[objectID].appearance_counter + 1

                self.current_objects[objectID].set_new_centroid(input_centroids[col], input_boxes[col], frame_id)
                self.current_objects[objectID].appearance_counter = appearance_counter
                self.disappeared_objects[objectID] = 0

                if appearance_counter == self.min_visibility:
                    self.current_objects_counter += 1
                    self.counter += 1
                elif appearance_counter > self.min_visibility:
                    self.current_objects[objectID].add_label(input_labels[col], input_confidences[col])

                used_rows.add(row)
                used_cols.add(col)

            unused_rows = set(range(0, D.shape[0])).difference(used_rows)
            unused_cols = set(range(0, D.shape[1])).difference(used_cols)

            if D.shape[0] >= D.shape[1]:
                for row in unused_rows:
                    objectID = object_ids[row]
                    self.disappeared_objects[objectID] += 1
                    self.current_objects[objectID].track_lost = True
                    # in this case maybe we can give some estimation
                    # self.current_objects[objectID].predict_new_centroid(frame_id)

                    if self.disappeared_objects[objectID] > self.max_disappeared:
                        self.deregister(objectID, time_data)

            else:
                for col in unused_cols:
                    self.register(input_centroids[col], input_boxes[col], time_data)
        self.avg_execution_sec = (self.avg_execution_sec + (time.time() - start_t)) / 2
        return list(self.current_objects.values())