Skip to content

A multi-object-tracking system which use tracking-by-detection method:yolo3/ssd(detection model)+sort/deep sort(tracking model)

Notifications You must be signed in to change notification settings

anhuipl2010/multi-object-tracking-yolo3-ssd-sort-deep-sort

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

multi-object-tracking:yolo3/ssd+sort/deep_sort

A multi-object-tracking system which use tracking-by-detection method:yolo3/ssd(detection model)+sort/deep sort(tracking model)

Introduction

The modification of this repository is based on the following repositories:

keras-yolo3

SSD-Tensorflow

sort

deep_sort

deep_sort_yolov3

centernet_tensorflow_wilderface_voc

preparation

  1. Download this repository.

  2. download YOLOv3 weights from YOLO website and put them in Repository_ROOT/yolo3_tf/.

  3. Convert the Darknet YOLO model to a Keras model.

python convert.py yolov3.cfg yolov3.weights ./model_data/yolo.h5
  1. download deep_sort pretrained weights from here and put them in Repository_ROOT/deep_sort/model_data/.
  2. Then generate file:mars-small128.pb.
cd Repository_ROOT/deep_sort/deep_sort/tools/
python freeze_model.py
  1. download SSD weights from here,uzip it and put them in Repository_ROOT/SSD_tf/model_data/.

  2. download file:yolo3_centernet_voc from here ,password:qqwx.Then put them in Repository_ROOT/yolo3_centernet_tf/model_data/.

  3. download MOT16 datasets from here ,uzip MOT16.zip,and put file: MOT16 in Repository_ROOT/

  4. Convert MOT16 image to video.

python Repository_ROOT/MOT16/convert_image_to_video.py

test

run SSD/YOLO3+sort:

python detect_and_sort_video.py
# just modify the parameter detection_mode="SSD"or"YOLO3" to choose different detection model

run SSD/YOLO3+deep_sort:

python detect_and_deep_sort_video.py
# just modify the parameter detection_mode="SSD"or"YOLO3" to choose different detection model

About

A multi-object-tracking system which use tracking-by-detection method:yolo3/ssd(detection model)+sort/deep sort(tracking model)

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Languages