pmj110119 / YOLOX_deepsort_tracker

using yolox+deepsort for object-tracking

YOLOX_deepsort_tracker

🎉 How to use

↳ Tracker example

from tracker import Tracker

tracker = Tracker()    # instantiate Tracker

cap = cv2.VideoCapture('test.mp4')  # open one video

while True:
    _, img = cap.read() # read frame from video
    if img is None:
       break
    
    img_visual, bbox = tracker.update(img)  # feed one frame and get result
    
    cv2.imshow('demo', img_visual)	# imshow
    cv2.waitKey(1)
    if cv2.getWindowProperty('demo', cv2.WND_PROP_AUTOSIZE) < 1:
        break

cap.release()
cv2.destroyAllWindows()

Tracker uses YOLOX as detector to get each target's boundingbox, and use deepsort to get every bbox's ID.

↳ Select specific category

If you just want to track only some specific categories, you can set by param filter_classes.

For example:

tracker = Tracker(filter_classes=['car','person']) 

↳ Detector example

If you don't need tracking and just want to use YOLOX for object-detection, you can use the class Detector to inference easliy .

For example:

from detector import Detector
import cv2
detector = Detector() # instantiate Detector

img = cv2.imread('YOLOX/assets/dog.jpg') 	# load image
result = detector.detect(img) 	# detect targets

img_visual = result['visual'] 	 # visualized image
cv2.imshow('detect', img_visual) # imshow
cv2.waitKey(0)

You can also get more information like raw_img/boudingbox/score/class_id from the result of detector.

🎨 Install

1. Clone the repository recursively:

   git clone --recurse-submodules https://github.com/pmj110119/YOLOX_deepsort_tracker.git

   If you already cloned and forgot to use --recurse-submodules you can run git submodule update --init(clone最新的YOLOX仓库)

2. Make sure that you fulfill all the requirements: Python 3.8 or later with all requirements.txt dependencies installed, including torch>=1.7. To install, run:

   pip install -r requirements.txt

⚡ Select a YOLOX family model

1. train your own model or just download pretrained models from https://github.com/Megvii-BaseDetection/YOLOX

   Model	size	mAPtest 0.5:0.95	Speed V100 (ms)	Params (M)	FLOPs (G)	weights
   YOLOX-s	640	39.6	9.8	9.0	26.8	onedrive/github
   YOLOX-m	640	46.4	12.3	25.3	73.8	onedrive/github
   YOLOX-l	640	50.0	14.5	54.2	155.6	onedrive/github
   YOLOX-x	640	51.2	17.3	99.1	281.9	onedrive/github
   YOLOX-Darknet53	640	47.4	11.1	63.7	185.3	onedrive/github

   Download yolox_s.pth to the folder weights , which is the default model path of Tracker.

2. You can also use other yolox models as detector,. For example:

   """
   YOLO family: yolox-s, yolox-m, yolox-l, yolox-x, yolox-tiny, yolox-nano, yolov3
   """
   # yolox-s example
   detector = Tracker(model='yolox-s', ckpt='./yolox_s.pth')
   # yolox-m example
   detector = Tracker(model='yolox-m', ckpt='./yolox_m.pth')

🌹 Run demo

python demo.py --path=test.mp4