679 lines
28 KiB
Python
679 lines
28 KiB
Python
# Ultralytics YOLO 🚀, AGPL-3.0 license
|
|
|
|
import numpy as np
|
|
import math
|
|
|
|
from ..utils import LOGGER
|
|
from ..utils.ops import xywh2ltwh, xywh2xyxy
|
|
from .basetrack import BaseTrack, TrackState
|
|
from .utils import matching
|
|
from .utils.kalman_filter import KalmanFilterXYAH
|
|
|
|
|
|
class STrack(BaseTrack):
|
|
"""
|
|
Single object tracking representation that uses Kalman filtering for state estimation.
|
|
|
|
This class is responsible for storing all the information regarding individual tracklets and performs state updates
|
|
and predictions based on Kalman filter.
|
|
|
|
Attributes:
|
|
shared_kalman (KalmanFilterXYAH): Shared Kalman filter that is used across all STrack instances for prediction.
|
|
_tlwh (np.ndarray): Private attribute to store top-left corner coordinates and width and height of bounding box.
|
|
kalman_filter (KalmanFilterXYAH): Instance of Kalman filter used for this particular object track.
|
|
mean (np.ndarray): Mean state estimate vector.
|
|
covariance (np.ndarray): Covariance of state estimate.
|
|
is_activated (bool): Boolean flag indicating if the track has been activated.
|
|
score (float): Confidence score of the track.
|
|
tracklet_len (int): Length of the tracklet.
|
|
cls (any): Class label for the object.
|
|
idx (int): Index or identifier for the object.
|
|
frame_id (int): Current frame ID.
|
|
start_frame (int): Frame where the object was first detected.
|
|
|
|
Methods:
|
|
predict(): Predict the next state of the object using Kalman filter.
|
|
multi_predict(stracks): Predict the next states for multiple tracks.
|
|
multi_gmc(stracks, H): Update multiple track states using a homography matrix.
|
|
activate(kalman_filter, frame_id): Activate a new tracklet.
|
|
re_activate(new_track, frame_id, new_id): Reactivate a previously lost tracklet.
|
|
update(new_track, frame_id): Update the state of a matched track.
|
|
convert_coords(tlwh): Convert bounding box to x-y-aspect-height format.
|
|
tlwh_to_xyah(tlwh): Convert tlwh bounding box to xyah format.
|
|
"""
|
|
|
|
shared_kalman = KalmanFilterXYAH()
|
|
|
|
def __init__(self, xywh, score, cls):
|
|
"""Initialize new STrack instance."""
|
|
super().__init__()
|
|
# xywh+idx or xywha+idx
|
|
assert len(xywh) in {5, 6}, f"expected 5 or 6 values but got {len(xywh)}"
|
|
self._tlwh = np.asarray(xywh2ltwh(xywh[:4]), dtype=np.float32)
|
|
self.kalman_filter = None
|
|
self.mean, self.covariance = None, None
|
|
self.is_activated = False
|
|
|
|
self.score = score
|
|
self.tracklet_len = 0
|
|
self.cls = cls
|
|
self.idx = xywh[-1]
|
|
self.angle = xywh[4] if len(xywh) == 6 else None
|
|
|
|
self.det = np.asarray(xywh2xyxy(xywh[:4]), dtype=np.float32)
|
|
|
|
self.recent_scores = [] # 用于存储最近10次的score
|
|
self.recent_scores.append(self.score)
|
|
self.tracklet_len_score = self.score
|
|
|
|
def predict(self):
|
|
"""Predicts mean and covariance using Kalman filter."""
|
|
mean_state = self.mean.copy()
|
|
if self.state != TrackState.Tracked:
|
|
mean_state[7] = 0
|
|
self.mean, self.covariance = self.kalman_filter.predict(mean_state, self.covariance)
|
|
|
|
def prior_predict(self):
|
|
"""Predicts mean and covariance using Kalman filter."""
|
|
mean_state = self.mean.copy()
|
|
if self.state != TrackState.Tracked:
|
|
return None
|
|
mean, covariance = self.kalman_filter.predict(mean_state, self.covariance)
|
|
return mean
|
|
|
|
@staticmethod
|
|
def multi_predict(stracks):
|
|
"""Perform multi-object predictive tracking using Kalman filter for given stracks."""
|
|
if len(stracks) <= 0:
|
|
return
|
|
multi_mean = np.asarray([st.mean.copy() for st in stracks])
|
|
multi_covariance = np.asarray([st.covariance for st in stracks])
|
|
for i, st in enumerate(stracks):
|
|
if st.state != TrackState.Tracked:
|
|
multi_mean[i][7] = 0
|
|
multi_mean, multi_covariance = STrack.shared_kalman.multi_predict(multi_mean, multi_covariance)
|
|
for i, (mean, cov) in enumerate(zip(multi_mean, multi_covariance)):
|
|
stracks[i].mean = mean
|
|
stracks[i].covariance = cov
|
|
|
|
@staticmethod
|
|
def multi_gmc(stracks, H=np.eye(2, 3)):
|
|
"""Update state tracks positions and covariances using a homography matrix."""
|
|
if len(stracks) > 0:
|
|
multi_mean = np.asarray([st.mean.copy() for st in stracks])
|
|
multi_covariance = np.asarray([st.covariance for st in stracks])
|
|
|
|
R = H[:2, :2]
|
|
R8x8 = np.kron(np.eye(4, dtype=float), R)
|
|
t = H[:2, 2]
|
|
|
|
for i, (mean, cov) in enumerate(zip(multi_mean, multi_covariance)):
|
|
mean = R8x8.dot(mean)
|
|
mean[:2] += t
|
|
cov = R8x8.dot(cov).dot(R8x8.transpose())
|
|
|
|
stracks[i].mean = mean
|
|
stracks[i].covariance = cov
|
|
|
|
def activate(self, kalman_filter, frame_id):
|
|
"""Start a new tracklet."""
|
|
self.kalman_filter = kalman_filter
|
|
self.track_id = self.next_id()
|
|
self.mean, self.covariance = self.kalman_filter.initiate(self.convert_coords(self._tlwh))
|
|
|
|
self.tracklet_len = 0
|
|
self.state = TrackState.Tracked
|
|
if frame_id == 1:
|
|
self.is_activated = True
|
|
self.frame_id = frame_id
|
|
self.start_frame = frame_id
|
|
|
|
def re_activate(self, new_track, frame_id, new_id=False):
|
|
"""Reactivates a previously lost track with a new detection."""
|
|
self.mean, self.covariance = self.kalman_filter.update(
|
|
self.mean, self.covariance, self.convert_coords(new_track.tlwh)
|
|
)
|
|
self.tracklet_len = 0
|
|
self.state = TrackState.Tracked
|
|
self.is_activated = True
|
|
self.frame_id = frame_id
|
|
if new_id:
|
|
self.track_id = self.next_id()
|
|
self.score = new_track.score
|
|
self.cls = new_track.cls
|
|
self.angle = new_track.angle
|
|
self.idx = new_track.idx
|
|
|
|
self.det = new_track.xyxy
|
|
|
|
self.recent_scores = [] # re clear
|
|
self.recent_scores.append(self.score)
|
|
self.tracklet_len_score = self.score
|
|
|
|
def update(self, new_track, frame_id):
|
|
"""
|
|
Update the state of a matched track.
|
|
|
|
Args:
|
|
new_track (STrack): The new track containing updated information.
|
|
frame_id (int): The ID of the current frame.
|
|
"""
|
|
self.frame_id = frame_id
|
|
self.tracklet_len += 1
|
|
|
|
new_tlwh = new_track.tlwh
|
|
self.mean, self.covariance = self.kalman_filter.update(
|
|
self.mean, self.covariance, self.convert_coords(new_tlwh)
|
|
)
|
|
self.state = TrackState.Tracked
|
|
self.is_activated = True
|
|
|
|
self.score = new_track.score
|
|
self.cls = new_track.cls
|
|
self.angle = new_track.angle
|
|
self.idx = new_track.idx
|
|
|
|
self.det = new_track.xyxy
|
|
|
|
self.recent_scores.append(self.score)
|
|
if len(self.recent_scores) > 10:
|
|
self.recent_scores.pop(0)
|
|
self.tracklet_len_score = sum(self.recent_scores)
|
|
|
|
def update_simi(self, new_track, frame_id):
|
|
"""
|
|
Update the state of a matched track.
|
|
|
|
Args:
|
|
new_track (STrack): The new track containing updated information.
|
|
frame_id (int): The ID of the current frame.
|
|
"""
|
|
self.frame_id = frame_id
|
|
self.tracklet_len += 1
|
|
|
|
# 认为存在突变
|
|
new_tlwh = new_track.tlwh
|
|
self.mean, self.covariance = self.kalman_filter.initiate(self.convert_coords(new_tlwh))
|
|
self.state = TrackState.Tracked
|
|
self.is_activated = True
|
|
|
|
self.score = new_track.score
|
|
self.cls = new_track.cls
|
|
self.angle = new_track.angle
|
|
self.idx = new_track.idx
|
|
|
|
self.det = new_track.xyxy
|
|
|
|
self.recent_scores.append(self.score)
|
|
if len(self.recent_scores) > 10:
|
|
self.recent_scores.pop(0)
|
|
self.tracklet_len_score = sum(self.recent_scores)
|
|
|
|
def convert_coords(self, tlwh):
|
|
"""Convert a bounding box's top-left-width-height format to its x-y-aspect-height equivalent."""
|
|
return self.tlwh_to_xyah(tlwh)
|
|
|
|
@property
|
|
def tlwh(self):
|
|
"""Get current position in bounding box format (top left x, top left y, width, height)."""
|
|
if self.mean is None:
|
|
return self._tlwh.copy()
|
|
ret = self.mean[:4].copy()
|
|
ret[2] *= ret[3]
|
|
ret[:2] -= ret[2:] / 2
|
|
return ret
|
|
|
|
@property
|
|
def xyxy(self):
|
|
"""Convert bounding box to format (min x, min y, max x, max y), i.e., (top left, bottom right)."""
|
|
ret = self.tlwh.copy()
|
|
ret[2:] += ret[:2]
|
|
return ret
|
|
|
|
@staticmethod
|
|
def tlwh_to_xyah(tlwh):
|
|
"""Convert bounding box to format (center x, center y, aspect ratio, height), where the aspect ratio is width /
|
|
height.
|
|
"""
|
|
ret = np.asarray(tlwh).copy()
|
|
ret[:2] += ret[2:] / 2
|
|
ret[2] /= ret[3]
|
|
return ret
|
|
|
|
@property
|
|
def xywh(self):
|
|
"""Get current position in bounding box format (center x, center y, width, height)."""
|
|
ret = np.asarray(self.tlwh).copy()
|
|
ret[:2] += ret[2:] / 2
|
|
return ret
|
|
|
|
@property
|
|
def xywha(self):
|
|
"""Get current position in bounding box format (center x, center y, width, height, angle)."""
|
|
if self.angle is None:
|
|
LOGGER.warning("WARNING ⚠️ `angle` attr not found, returning `xywh` instead.")
|
|
return self.xywh
|
|
return np.concatenate([self.xywh, self.angle[None]])
|
|
|
|
@property
|
|
def result(self):
|
|
"""Get current tracking results."""
|
|
coords = self.xyxy if self.angle is None else self.xywha
|
|
return coords.tolist() + [self.track_id, self.score, self.cls, self.idx]
|
|
|
|
def __repr__(self):
|
|
"""Return a string representation of the SOTTracker object with start and end frames and track ID."""
|
|
return f"OT_{self.track_id}_({self.start_frame}-{self.end_frame})"
|
|
|
|
|
|
class SOTTracker:
|
|
"""
|
|
SOTTracker: A tracking algorithm built on top of YOLOv8 for object detection and tracking.
|
|
|
|
The class is responsible for initializing, updating, and managing the tracks for detected objects in a video
|
|
sequence. It maintains the state of tracked, lost, and removed tracks over frames, utilizes Kalman filtering for
|
|
predicting the new object locations, and performs data association.
|
|
|
|
Attributes:
|
|
tracked_stracks (list[STrack]): List of successfully activated tracks.
|
|
lost_stracks (list[STrack]): List of lost tracks.
|
|
removed_stracks (list[STrack]): List of removed tracks.
|
|
frame_id (int): The current frame ID.
|
|
args (namespace): Command-line arguments.
|
|
max_time_lost (int): The maximum frames for a track to be considered as 'lost'.
|
|
kalman_filter (object): Kalman Filter object.
|
|
|
|
Methods:
|
|
update(results, img=None): Updates object tracker with new detections.
|
|
get_kalmanfilter(): Returns a Kalman filter object for tracking bounding boxes.
|
|
init_track(dets, scores, cls, img=None): Initialize object tracking with detections.
|
|
get_dists(tracks, detections): Calculates the distance between tracks and detections.
|
|
multi_predict(tracks): Predicts the location of tracks.
|
|
reset_id(): Resets the ID counter of STrack.
|
|
joint_stracks(tlista, tlistb): Combines two lists of stracks.
|
|
sub_stracks(tlista, tlistb): Filters out the stracks present in the second list from the first list.
|
|
remove_duplicate_stracks(stracksa, stracksb): Removes duplicate stracks based on IoU.
|
|
"""
|
|
|
|
def __init__(self, args, frame_rate=30):
|
|
"""Initialize a YOLOv8 object to track objects with given arguments and frame rate."""
|
|
self.tracked_stracks = [] # type: list[STrack]
|
|
self.lost_stracks = [] # type: list[STrack]
|
|
self.removed_stracks = [] # type: list[STrack]
|
|
|
|
self.frame_id = 0
|
|
self.args = args
|
|
self.max_time_lost = int(frame_rate / 30.0 * args.track_buffer)
|
|
self.kalman_filter = self.get_kalmanfilter()
|
|
self.reset_id()
|
|
self.prior_strack = None
|
|
|
|
def update(self, results, img=None):
|
|
"""Updates object tracker with new detections and returns tracked object bounding boxes."""
|
|
self.frame_id += 1
|
|
self.prior_strack = None
|
|
activated_stracks = []
|
|
refind_stracks = []
|
|
lost_stracks = []
|
|
removed_stracks = []
|
|
|
|
scores = results.conf
|
|
bboxes = results.xywhr if hasattr(results, "xywhr") else results.xywh
|
|
# Add index
|
|
bboxes = np.concatenate([bboxes, np.arange(len(bboxes)).reshape(-1, 1)], axis=-1)
|
|
cls = results.cls
|
|
|
|
# if scores[0] < 0.1: # note:当box为0时IOU为1
|
|
# return np.asarray([])
|
|
|
|
# Add newly detected tracklets to tracked_stracks
|
|
unconfirmed = []
|
|
tracked_stracks = [] # type: list[STrack]
|
|
for track in self.tracked_stracks:
|
|
if not track.is_activated:
|
|
unconfirmed.append(track)
|
|
else:
|
|
tracked_stracks.append(track)
|
|
strack_pool = self.joint_stracks(tracked_stracks, self.lost_stracks)
|
|
# Predict the current location with KF
|
|
self.multi_predict(strack_pool) # important for correct state
|
|
if hasattr(self, "gmc") and img is not None:
|
|
warp = self.gmc.apply(img, dets)
|
|
STrack.multi_gmc(strack_pool, warp)
|
|
STrack.multi_gmc(unconfirmed, warp)
|
|
|
|
# step 000 for sot
|
|
prior_strack = None
|
|
prior_update_id = -1
|
|
candidate_det = None
|
|
len_max = 0
|
|
# 寻找优势目标
|
|
# 优先在激活目标中寻找
|
|
for track in self.tracked_stracks:
|
|
if track.tracklet_len_score > max(len_max, 5):
|
|
prior_strack = track
|
|
len_max = track.tracklet_len_score
|
|
# 没有则在短时丢失目标中寻找
|
|
if prior_strack is None:
|
|
for track in self.lost_stracks:
|
|
if track.tracklet_len_score > max(len_max, 5):
|
|
prior_strack = track
|
|
len_max = track.tracklet_len_score
|
|
# 存在优势目标后进行匹配
|
|
if prior_strack is not None:
|
|
sot_detections = self.init_track(bboxes, scores, cls, img)
|
|
simi_dists = self.get_simi_dists([prior_strack], sot_detections)
|
|
iou_dists = matching.iou_distance([prior_strack], sot_detections)
|
|
# idet = np.argmax(iou_dists)
|
|
# if iou_dists[0, idet] > 0.5: # note:当box为0时IOU为1
|
|
# prior_strack.update(sot_detections[idet], self.frame_id)
|
|
# activated_stracks.append(prior_strack)
|
|
# prior_update_id = idet
|
|
# elif simi_dists.max() > 0.6: # IOU计算替换:
|
|
# # 置信度较高则直接更新
|
|
# if sot_detections[idet].score > 0.3:
|
|
# prior_strack.update_simi(sot_detections[idet], self.frame_id)
|
|
# activated_stracks.append(prior_strack)
|
|
# prior_update_id = idet
|
|
# # 置信度较低则采用更严格匹配
|
|
# else:
|
|
# candidate_det = sot_detections[idet] # track_id
|
|
# candidate_det.track_id = prior_strack.track_id
|
|
idet = np.argmax(simi_dists)
|
|
if simi_dists.max() > 0.6: # IOU计算替换
|
|
# 置信度较高则直接更新
|
|
if iou_dists[0, idet] > 0.5:
|
|
prior_strack.update(sot_detections[idet], self.frame_id)
|
|
activated_stracks.append(prior_strack)
|
|
prior_update_id = idet
|
|
else:
|
|
if sot_detections[idet].score > 0.2: # 0.3
|
|
prior_strack.update_simi(sot_detections[idet], self.frame_id)
|
|
activated_stracks.append(prior_strack)
|
|
prior_update_id = idet
|
|
# 置信度较低则采用更严格匹配
|
|
else:
|
|
candidate_det = sot_detections[idet] # track_id
|
|
candidate_det.track_id = prior_strack.track_id
|
|
|
|
if prior_update_id != -1:
|
|
scores[idet] = 0 # for sot
|
|
remain_inds = scores >= self.args.track_high_thresh
|
|
inds_low = scores > self.args.track_low_thresh
|
|
inds_high = scores < self.args.track_high_thresh
|
|
|
|
inds_second = inds_low & inds_high
|
|
dets_second = bboxes[inds_second]
|
|
dets = bboxes[remain_inds]
|
|
scores_keep = scores[remain_inds]
|
|
scores_second = scores[inds_second]
|
|
cls_keep = cls[remain_inds]
|
|
cls_second = cls[inds_second]
|
|
|
|
# Step 2: First association, with high score detection boxes
|
|
detections = self.init_track(dets, scores_keep, cls_keep, img)
|
|
if prior_update_id != -1:
|
|
strack_pool = self.sub_stracks(strack_pool, [prior_strack]) # for sot
|
|
dists = self.get_dists(strack_pool, detections)
|
|
matches, u_track, u_detection = matching.linear_assignment(dists, thresh=self.args.match_thresh)
|
|
|
|
for itracked, idet in matches:
|
|
track = strack_pool[itracked]
|
|
det = detections[idet]
|
|
if track.state == TrackState.Tracked:
|
|
track.update(det, self.frame_id)
|
|
activated_stracks.append(track)
|
|
else:
|
|
track.re_activate(det, self.frame_id, new_id=False)
|
|
refind_stracks.append(track)
|
|
# Step 3: Second association, with low score detection boxes association the untrack to the low score detections
|
|
detections_second = self.init_track(dets_second, scores_second, cls_second, img)
|
|
r_tracked_stracks = [strack_pool[i] for i in u_track if strack_pool[i].state == TrackState.Tracked]
|
|
# TODO
|
|
dists = matching.iou_distance(r_tracked_stracks, detections_second)
|
|
matches, u_track, u_detection_second = matching.linear_assignment(dists, thresh=0.5)
|
|
for itracked, idet in matches:
|
|
track = r_tracked_stracks[itracked]
|
|
det = detections_second[idet]
|
|
if track.state == TrackState.Tracked:
|
|
track.update(det, self.frame_id)
|
|
activated_stracks.append(track)
|
|
else:
|
|
track.re_activate(det, self.frame_id, new_id=False)
|
|
refind_stracks.append(track)
|
|
|
|
for it in u_track:
|
|
track = r_tracked_stracks[it]
|
|
if track.state != TrackState.Lost:
|
|
track.mark_lost()
|
|
lost_stracks.append(track)
|
|
# Deal with unconfirmed tracks, usually tracks with only one beginning frame
|
|
detections = [detections[i] for i in u_detection]
|
|
dists = self.get_dists(unconfirmed, detections)
|
|
matches, u_unconfirmed, u_detection = matching.linear_assignment(dists, thresh=0.7)
|
|
for itracked, idet in matches:
|
|
unconfirmed[itracked].update(detections[idet], self.frame_id)
|
|
activated_stracks.append(unconfirmed[itracked])
|
|
for it in u_unconfirmed:
|
|
track = unconfirmed[it]
|
|
track.mark_removed()
|
|
removed_stracks.append(track)
|
|
# Step 4: Init new stracks
|
|
for inew in u_detection:
|
|
track = detections[inew]
|
|
if track.score < self.args.new_track_thresh:
|
|
continue
|
|
track.activate(self.kalman_filter, self.frame_id)
|
|
activated_stracks.append(track)
|
|
# Step 5: Update state
|
|
for track in self.lost_stracks:
|
|
if self.frame_id - track.end_frame > self.max_time_lost:
|
|
track.mark_removed()
|
|
removed_stracks.append(track)
|
|
|
|
self.tracked_stracks = [t for t in self.tracked_stracks if t.state == TrackState.Tracked]
|
|
self.tracked_stracks = self.joint_stracks(self.tracked_stracks, activated_stracks)
|
|
self.tracked_stracks = self.joint_stracks(self.tracked_stracks, refind_stracks)
|
|
self.lost_stracks = self.sub_stracks(self.lost_stracks, self.tracked_stracks)
|
|
self.lost_stracks.extend(lost_stracks)
|
|
self.lost_stracks = self.sub_stracks(self.lost_stracks, self.removed_stracks)
|
|
self.tracked_stracks, self.lost_stracks = self.remove_duplicate_stracks(self.tracked_stracks, self.lost_stracks)
|
|
self.removed_stracks.extend(removed_stracks)
|
|
if len(self.removed_stracks) > 1000:
|
|
self.removed_stracks = self.removed_stracks[-999:] # clip remove stracks to 1000 maximum
|
|
|
|
# 当多个目标时：只输出优先目标
|
|
# 当没有可靠目标但有检测时：输出simi最相似的目标
|
|
prior_strack = None
|
|
len_max = -1
|
|
for track in self.tracked_stracks:
|
|
if track.state == TrackState.Tracked and track.tracklet_len_score > len_max:
|
|
prior_strack = track
|
|
len_max = track.tracklet_len_score
|
|
# 初始化点置信度将为0.25，按轨迹长度进行置信度分级，并采用轨迹加权以减轻一直误检影响
|
|
score_max = 0
|
|
if prior_strack is not None and prior_strack.tracklet_len_score < 1:
|
|
for track in self.tracked_stracks: #选出置信度最高
|
|
if track.state == TrackState.Tracked and track.score > score_max:
|
|
prior_strack = track
|
|
score_max = track.score
|
|
if prior_strack.score < 0.6:
|
|
prior_strack = None
|
|
if prior_strack is not None:
|
|
self.prior_strack = prior_strack
|
|
return np.asarray([prior_strack.det.tolist()+prior_strack.result[-4:]], dtype=np.float32) # result为滤波调整后
|
|
elif candidate_det is not None:
|
|
return np.asarray([candidate_det.result], dtype=np.float32)
|
|
else:
|
|
# valid_indices = np.where(scores > 0)[0]
|
|
# if valid_indices.size > 0:
|
|
# max_idx = valid_indices[np.argmax(scores[valid_indices])]
|
|
# out_det = STrack(bboxes[max_idx], scores[max_idx], cls[max_idx])
|
|
# return np.asarray([out_det.result], dtype=np.float32)
|
|
return np.asarray([])
|
|
|
|
# return np.asarray([x.result for x in self.tracked_stracks if x.is_activated], dtype=np.float32)
|
|
|
|
def prior_predict(self):
|
|
"""Returns a Kalman filter object for tracking bounding boxes."""
|
|
if self.prior_strack is not None:
|
|
return self.prior_strack.prior_predict()
|
|
return None
|
|
|
|
def multi_gmc(self, warp):
|
|
STrack.multi_gmc(self.tracked_stracks, warp)
|
|
STrack.multi_gmc(self.lost_stracks, warp)
|
|
|
|
def get_kalmanfilter(self):
|
|
"""Returns a Kalman filter object for tracking bounding boxes."""
|
|
return KalmanFilterXYAH()
|
|
|
|
def init_track(self, dets, scores, cls, img=None):
|
|
"""Initialize object tracking with detections and scores using STrack algorithm."""
|
|
return [STrack(xyxy, s, c) for (xyxy, s, c) in zip(dets, scores, cls)] if len(dets) else [] # detections
|
|
|
|
def get_dists(self, tracks, detections):
|
|
"""Calculates the distance between tracks and detections using IoU and fuses scores."""
|
|
dists = matching.iou_distance(tracks, detections)
|
|
# TODO: mot20
|
|
# if not self.args.mot20:
|
|
dists = matching.fuse_score(dists, detections)
|
|
return dists
|
|
|
|
def get_simi_dists(self, tracks, detections):
|
|
"""Calculates the distance between tracks and detections using IoU and fuses scores."""
|
|
dists = self.simi_distance(tracks, detections)
|
|
return dists
|
|
|
|
def multi_predict(self, tracks):
|
|
"""Returns the predicted tracks using the YOLOv8 network."""
|
|
STrack.multi_predict(tracks)
|
|
|
|
@staticmethod
|
|
def reset_id():
|
|
"""Resets the ID counter of STrack."""
|
|
STrack.reset_id()
|
|
|
|
def reset(self):
|
|
"""Reset tracker."""
|
|
self.tracked_stracks = [] # type: list[STrack]
|
|
self.lost_stracks = [] # type: list[STrack]
|
|
self.removed_stracks = [] # type: list[STrack]
|
|
self.frame_id = 0
|
|
self.kalman_filter = self.get_kalmanfilter()
|
|
self.reset_id()
|
|
|
|
@staticmethod
|
|
def joint_stracks(tlista, tlistb):
|
|
"""Combine two lists of stracks into a single one."""
|
|
exists = {}
|
|
res = []
|
|
for t in tlista:
|
|
exists[t.track_id] = 1
|
|
res.append(t)
|
|
for t in tlistb:
|
|
tid = t.track_id
|
|
if not exists.get(tid, 0):
|
|
exists[tid] = 1
|
|
res.append(t)
|
|
return res
|
|
|
|
@staticmethod
|
|
def sub_stracks(tlista, tlistb):
|
|
"""DEPRECATED CODE in https://github.com/ultralytics/ultralytics/pull/1890/
|
|
stracks = {t.track_id: t for t in tlista}
|
|
for t in tlistb:
|
|
tid = t.track_id
|
|
if stracks.get(tid, 0):
|
|
del stracks[tid]
|
|
return list(stracks.values())
|
|
"""
|
|
track_ids_b = {t.track_id for t in tlistb}
|
|
return [t for t in tlista if t.track_id not in track_ids_b]
|
|
|
|
@staticmethod
|
|
def remove_duplicate_stracks(stracksa, stracksb):
|
|
"""Remove duplicate stracks with non-maximum IoU distance."""
|
|
pdist = matching.iou_distance(stracksa, stracksb)
|
|
pairs = np.where(pdist < 0.15)
|
|
dupa, dupb = [], []
|
|
for p, q in zip(*pairs):
|
|
timep = stracksa[p].frame_id - stracksa[p].start_frame
|
|
timeq = stracksb[q].frame_id - stracksb[q].start_frame
|
|
if timep > timeq:
|
|
dupb.append(q)
|
|
else:
|
|
dupa.append(p)
|
|
resa = [t for i, t in enumerate(stracksa) if i not in dupa]
|
|
resb = [t for i, t in enumerate(stracksb) if i not in dupb]
|
|
return resa, resb
|
|
|
|
@staticmethod
|
|
def simi_distance(atracks: list, btracks: list) -> np.ndarray:
|
|
"""
|
|
Compute cost based on Intersection over Union (IoU) between tracks.
|
|
|
|
Args:
|
|
atracks (list[STrack] | list[np.ndarray]): List of tracks 'a' or bounding boxes.
|
|
btracks (list[STrack] | list[np.ndarray]): List of tracks 'b' or bounding boxes.
|
|
|
|
Returns:
|
|
(np.ndarray): Cost matrix computed based on IoU.
|
|
"""
|
|
|
|
if atracks and isinstance(atracks[0], np.ndarray) or btracks and isinstance(btracks[0], np.ndarray):
|
|
atlbrs = atracks
|
|
btlbrs = btracks
|
|
else:
|
|
atlbrs = [track.xywha if track.angle is not None else track.xyxy for track in atracks]
|
|
btlbrs = [track.xywha if track.angle is not None else track.xyxy for track in btracks]
|
|
|
|
simi = np.zeros((len(atlbrs), len(btlbrs)), dtype=np.float32)
|
|
if len(atlbrs) and len(btlbrs):
|
|
simi = SOTTracker.bbox_simi(
|
|
np.ascontiguousarray(atlbrs, dtype=np.float32),
|
|
np.ascontiguousarray(btlbrs, dtype=np.float32),
|
|
)
|
|
return 1 - simi # cost matrix
|
|
|
|
@staticmethod
|
|
def bbox_simi(box1, box2, eps=1e-7):
|
|
"""
|
|
Calculate the intersection over box2 area given box1 and box2. Boxes are in x1y1x2y2 format.
|
|
|
|
Args:
|
|
box1 (np.ndarray): A numpy array of shape (n, 4) representing n bounding boxes.
|
|
box2 (np.ndarray): A numpy array of shape (m, 4) representing m bounding boxes.
|
|
iou (bool): Calculate the standard IoU if True else return inter_area/box2_area.
|
|
eps (float, optional): A small value to avoid division by zero. Defaults to 1e-7.
|
|
|
|
Returns:
|
|
(np.ndarray): A numpy array of shape (n, m) representing the intersection over box2 area.
|
|
"""
|
|
|
|
# Get the coordinates of bounding boxes
|
|
b1_x1, b1_y1, b1_x2, b1_y2 = box1.T
|
|
b2_x1, b2_y1, b2_x2, b2_y2 = box2.T
|
|
cx1 = (b1_x1 + b1_x2) / 2
|
|
cy1 = (b1_y1 + b1_y2) / 2
|
|
cx2 = (b2_x1 + b2_x2) / 2
|
|
cy2 = (b2_y1 + b2_y2) / 2
|
|
w1 = b1_x2 - b1_x1
|
|
h1 = b1_y2 - b1_y1
|
|
w2 = b2_x2 - b2_x1
|
|
h2 = b2_y2 - b2_y1
|
|
cw = np.maximum(b1_x2,b2_x2) - np.minimum(b1_x1,b2_x1) # convex (smallest enclosing box) width
|
|
ch = np.maximum(b1_y2,b2_y2) - np.minimum(b1_y1,b2_y1) # convex height
|
|
|
|
# shape = abs(w1*h1-w2*h2) + abs(w1*h2-w2*h1)
|
|
# dis = (cx1 - cx2)**2 + (cy1 - cy2)**2
|
|
|
|
shape = 0.5*np.abs(w1*h1-w2*h2)/(np.maximum(w1*h1,w2*h2)+eps) + 0.5*np.abs(w1*h2-w2*h1)/(np.maximum(w1*h2,w2*h1)+eps)
|
|
dis = np.sqrt((cx1 - cx2)**2 + (cy1 - cy2)**2) / (np.sqrt(cw**2 + ch**2) + eps)
|
|
|
|
simi = shape * 0.8 + dis * 0.2
|
|
|
|
# if shape.shape[0] > 1:
|
|
# print(simi)
|
|
|
|
return simi |