^j t d dl mZ d dlZd dlmZmZ d dlmZ d dlm Z d dl
Z d dl m
Z
d dlmZ G d d Zy)
) annotationsN) defaultdictdeque)deepcopy)cast)
Detections)SupervisionWarningsc 2 e Zd ZdZdddZd dZd
dZd dZy) DetectionsSmootheru
A utility class for smoothing detections over multiple frames in video tracking.
It maintains a history of detections for each track and provides smoothed
predictions based on these histories.
!!! warning
- `DetectionsSmoother` requires the `tracker_id` for each detection. Refer to
[Roboflow Trackers](/latest/trackers/) for
information on integrating tracking into your inference pipeline.
- This class is not compatible with segmentation models.
- When detections in a frame disagree on confidence presence — some tracks
carry confidence scores and others do not — `confidence` is set to `None`
for all smoothed detections in that frame.
Example:
```pycon
>>> import numpy as np
>>> import supervision as sv
>>> smoother = sv.DetectionsSmoother(length=3)
>>> detections_1 = sv.Detections(
... xyxy=np.array([[0, 0, 10, 10]]),
... confidence=np.array([0.5]),
... tracker_id=np.array([1])
... )
>>> detections_2 = sv.Detections(
... xyxy=np.array([[2, 2, 12, 12]]),
... confidence=np.array([0.7]),
... tracker_id=np.array([1])
... )
>>> smoothed = smoother.update_with_detections(detections_1)
>>> smoothed.xyxy
array([[ 0., 0., 10., 10.]])
>>> smoothed = smoother.update_with_detections(detections_2)
>>> smoothed.xyxy
array([[ 1., 1., 11., 11.]])
>>> smoothed.confidence
array([0.6])
```
```python
import supervision as sv
from ultralytics import YOLO
video_info = sv.VideoInfo.from_video_path(video_path="")
frame_generator = sv.get_video_frames_generator(
source_path="")
model = YOLO("")
tracker = sv.ByteTrack(frame_rate=video_info.fps)
smoother = sv.DetectionsSmoother()
box_annotator = sv.BoxAnnotator()
with sv.VideoSink("", video_info=video_info) as sink:
for frame in frame_generator:
result = model(frame)[0]
detections = sv.Detections.from_ultralytics(result)
detections = tracker.update_with_detections(detections)
detections = smoother.update_with_detections(detections)
annotated_frame = box_annotator.annotate(frame.copy(), detections)
sink.write_frame(annotated_frame)
```
c , t fd | _ y)z
Args:
length: The maximum number of frames to consider for smoothing
detections. Defaults to 5.
c t S )N)maxlen)r )lengths o/var/www/ramen.bs-engineer-server.com/venv/lib/python3.12/site-packages/supervision/detection/tools/smoother.pyz-DetectionsSmoother.__init__..a s E( N)r tracks)selfr s `r __init__zDetectionsSmoother.__init__Z s CN (C
r c |j t j dt |S t t | D ]K }|j | }t
| }| j | j t t || M | j j D ]/ }||j vs| j | j d 1 t | j j D ]8 }t | j | D cg c] }|du c} s,| j |= : | j S c c}w )z
Updates the smoother with a new set of detections from a frame.
Args:
detections: The detections to add to the smoother.
NzSmoothing skipped. DetectionsSmoother requires tracker_id. Refer to https://supervision.roboflow.com/latest/trackers for more information.)category)
tracker_idwarningswarnr rangelenintr appendr r keyslistallget_smoothed_detections)r
detections
detection_idxtracker_id_valuer track_idds r update_with_detectionsz)DetectionsSmoother.update_with_detectionsd s/ ( MM -
"3z? 3 XM ) 4 4] C -.J KK
# * *4
J}KK) * ++-- >s E
c | j j |d }|y|D cg c] }|| }}t | dk( ryt |d }t j
|D cg c] }|j c}d |_ |D cg c] }|j |j }}|r1t j
t j | d |_ |S d|_ |S c c}w c c}w c c}w )aE Return the smoothed `Detections` for a single track.
Averages `xyxy` over all valid (non-`None`) frames in the track window.
`confidence` is averaged only over frames that carry it; frames with
`confidence=None` are excluded. Returns `None` when the track is unknown
or its entire window is empty.
Args:
track_id: The tracker ID whose smoothed detection to retrieve.
Returns:
Smoothed `Detections` for the track, or `None` if the track is
unknown or all frames in its window are empty.
Nr )axis) r getr r npmeanxyxy
confidencearray)r r& trackr' validret confidencess r get_trackz DetectionsSmoother.get_track s $/ =.3"Eq}1"E"E u:?uQx 77E2qAFF2; .3Oall6Nq||O OCN+!6Q?
UY
#F
3 Ps C%C%"C*C/C/c @ g }| j D ]' }| j | }||j | ) |r t d |D r|D ] }d |_ t j | }t | dk( r t j g t |_ |S )Nc 3 8 K | ] }|j d u y w)N)r/ ).0r' s r z=DetectionsSmoother.get_smoothed_detections.. s %Wqalld&:%Ws r )dtype) r r5 r anyr/ r merger r, r0 r r )r tracked_detectionsr& r1 r' r# s r r" z*DetectionsSmoother.get_smoothed_detections s 1HNN8,E "))%0 1 #%WDV%W"W'
$ #
$ %%&89
z?a $&HHRs$;J !r N) )r r returnNone)r# r r? r )r& r r? zDetections | None)r? r )__name__
__module__ __qualname____doc__r r( r5 r" r r r r s IV
.B Br r )
__future__r r collectionsr r copyr typingr numpyr, supervision.detection.corer supervision.utils.internalr r rE r r rM s) " * 1 :j jr