Fix track id remains unchanged for different bounding boxes, missing attribute error

yolox/tracker/basetrack.py

@@ -10,31 +10,35 @@ class TrackState(object):
 
 
 class BaseTrack(object):
-    _count = 0
+    def __init__(self, count_gen=None):
+        self._count = 0
+        self.count_gen = count_gen
 
-    track_id = 0
-    is_activated = False
-    state = TrackState.New
+        self.track_id = 0
+        self.is_activated = False
+        self.state = TrackState.New
 
-    history = OrderedDict()
-    features = []
-    curr_feature = None
-    score = 0
-    start_frame = 0
-    frame_id = 0
-    time_since_update = 0
+        self.history = OrderedDict()
+        self.features = []
+        self.curr_feature = None
+        self.score = 0
+        self.start_frame = 0
+        self.frame_id = 0
+        self.time_since_update = 0
 
-    # multi-camera
-    location = (np.inf, np.inf)
+        # multi-camera
+        self.location = (np.inf, np.inf)
 
     @property
     def end_frame(self):
         return self.frame_id
 
-    @staticmethod
-    def next_id():
-        BaseTrack._count += 1
-        return BaseTrack._count
+    def next_id(self):
+        if self.count_gen:
+            self._count = self.count_gen.__next__()
+        else:
+            self._count += 1
+        return self._count
 
     def activate(self, *args):
         raise NotImplementedError

yolox/tracker/byte_tracker.py

@@ -5,15 +5,16 @@
 import copy
 import torch
 import torch.nn.functional as F
+import itertools
 
 from .kalman_filter import KalmanFilter
 from yolox.tracker import matching
 from .basetrack import BaseTrack, TrackState
 
 class STrack(BaseTrack):
     shared_kalman = KalmanFilter()
-    def __init__(self, tlwh, score):
-
+    def __init__(self, tlwh, score, count_gen=None):
+        super().__init__(count_gen)
         # wait activate
         self._tlwh = np.asarray(tlwh, dtype=np.float)
         self.kalman_filter = None
@@ -155,6 +156,7 @@ def __init__(self, args, frame_rate=30):
         self.buffer_size = int(frame_rate / 30.0 * args.track_buffer)
         self.max_time_lost = self.buffer_size
         self.kalman_filter = KalmanFilter()
+        self.track_count_gen = itertools.count(start=1)
 
     def update(self, output_results, img_info, img_size):
         self.frame_id += 1
@@ -186,7 +188,7 @@ def update(self, output_results, img_info, img_size):
 
         if len(dets) > 0:
             '''Detections'''
-            detections = [STrack(STrack.tlbr_to_tlwh(tlbr), s) for
+            detections = [STrack(STrack.tlbr_to_tlwh(tlbr), s, self.track_count_gen) for
                           (tlbr, s) in zip(dets, scores_keep)]
         else:
             detections = []
@@ -223,7 +225,7 @@ def update(self, output_results, img_info, img_size):
         # association the untrack to the low score detections
         if len(dets_second) > 0:
             '''Detections'''
-            detections_second = [STrack(STrack.tlbr_to_tlwh(tlbr), s) for
+            detections_second = [STrack(STrack.tlbr_to_tlwh(tlbr), s, self.track_count_gen) for
                           (tlbr, s) in zip(dets_second, scores_second)]
         else:
             detections_second = []