视频图matlab代码-Tracking_by_detection:系统使用yolov3/KCF&DSST/KF来检测,跟踪和估

  • E9_765208
    了解作者
  • 197.2KB
    文件大小
  • zip
    文件格式
  • 0
    收藏次数
  • VIP专享
    资源类型
  • 0
    下载次数
  • 2022-04-13 04:27
    上传日期
视频图matlab代码 Tracking_by_detection A Python3-based system use yolov3/KCF&DSST/KF to detect,tracking and estimate the target(people and cars) location. 这是我的本科毕设项目,主要通过YOLOv3识别,利用KCF-DSST-APCE完成抗遮挡的尺度变化跟踪,并用卡尔曼滤波估计位置。 视频效果/video : Test environment Intel Core i5-8300H Nvidia GTX 1050Ti Ubuntu 18.04LTS Python3 OPENCV3.4.4 Requirements CUDA9.0 CUDNN7 Python3 OPENCV Numpy Numba Code structure yolo.py:detect fhog.py:give fhog feature tracker.py:track run.py:basic framework run2.py:add more visualization
Tracking_by_detection-master.zip
  • Tracking_by_detection-master
  • model_data
  • yolo_anchors.txt
    77B
  • coco_classes.txt
    705B
  • yolo3_object.names
    63B
  • yolov3.cfg
    8.9KB
  • obj.txt
    63B
  • voc_classes.txt
    155B
  • __pycache__
  • fhog.func1-16.py36m.nbi
    1.1KB
  • fhog.func4-112.py36m.nbi
    1.1KB
  • fhog.func3-75.py36m.nbi
    1KB
  • fhog.func2-54.py36m.1.nbc
    87.4KB
  • fhog.func3-75.py36m.1.nbc
    153.8KB
  • tracker.cpython-36.pyc
    15.1KB
  • fhog.func1-16.py36m.1.nbc
    54.9KB
  • fhog.cpython-36.pyc
    6.3KB
  • fhog.func4-112.py36m.1.nbc
    44.3KB
  • yolo.cpython-36.pyc
    4.7KB
  • fhog.func2-54.py36m.nbi
    1.3KB
  • yolo3
  • __pycache__
  • utils.cpython-36.pyc
    1.4KB
  • model.cpython-36.pyc
    11.2KB
  • utils.py
    1004B
  • model.py
    13.9KB
  • fhog.py
    12.3KB
  • run2.py
    12.1KB
  • run.py
    7.6KB
  • yolo.py
    6.2KB
  • tracker.py
    25.1KB
  • LICENSE
    11.1KB
  • README.md
    3.5KB
内容介绍
# Tracking_by_detection A Python3-based system use yolov3/KCF&DSST/KF to detect,tracking and estimate the target(people and cars) location. 这是我的本科毕设项目,主要通过YOLOv3识别,利用KCF-DSST-APCE完成抗遮挡的尺度变化跟踪,并用卡尔曼滤波估计位置。 # 视频效果/video : https://www.bilibili.com/video/bv1Xa4y1Y7Ue # Test environment Intel Core i5-8300H CPU@2.30GHz Nvidia GTX 1050Ti Ubuntu 18.04LTS Python3 OPENCV3.4.4 # Requirements CUDA9.0 CUDNN7 Python3 OPENCV Numpy Numba # Code structure yolo.py:detect fhog.py:give fhog feature tracker.py:track run.py:basic framework run2.py:add more visualization work for some video # Cites and Others' codes' citation @article{redmon2018yolov3, title={Yolov3: An incremental improvement}, author={Redmon, Joseph and Farhadi, Ali}, journal={arXiv preprint arXiv:1804.02767}, year={2018} } @article{henriques2014high, title={High-speed tracking with kernelized correlation filters}, author={Henriques, Jo{\~a}o F and Caseiro, Rui and Martins, Pedro and Batista, Jorge}, journal={IEEE transactions on pattern analysis and machine intelligence}, volume={37}, number={3}, pages={583--596}, year={2014}, publisher={IEEE} } @inproceedings{danelljan2014accurate, title={Accurate scale estimation for robust visual tracking}, author={Danelljan, Martin and H{\"a}ger, Gustav and Khan, Fahad and Felsberg, Michael}, booktitle={British Machine Vision Conference, Nottingham, September 1-5, 2014}, year={2014}, organization={BMVA Press} } @inproceedings{wang2017large, title={Large margin object tracking with circulant feature maps}, author={Wang, Mengmeng and Liu, Yong and Huang, Zeyi}, booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition}, pages={4021--4029}, year={2017} } The YOLOv3 part is based on yehengcheng's work @ https://github.com/yehengchen/Object-Detection-and-Tracking The KCF and DSST part is based on ryanfwy's work @ https://github.com/ryanfwy/KCF-DSST-py # Install and use ## First you should clone this repository and get yolo weight: > git clone https://github.com/sjtuzyz/Tracking_by_detection.git go to yehengcheng's work to get the yolo weight or learn how to transfer it. ## Then change some part of code to slove your task ### Go to yolo.py line 122 > if predicted_class != 'person': The class should be changed to person or car or whatever you want.About the yolo network's training and convert,please check yehengcheng's work. ### Go to tracker.py line 463 ```python if (APCE>0.12):#25 oc1.mp4 0.02 oc2.mp4 x = self.getFeatures(image, 0, 1.0) self.train(x, self.interp_factor) #x = self.getFeatures(image, 0, 1.0) #self.train(x, self.interp_factor) ``` If you want to use APCE to let the system more robust to occlusions ,just change 0.12 to a threshold of your test video, which will be print in the terminal. ### Go to run.py(or run2.py and find where to change by yourself)line 35 > cap = cv2.VideoCapture('./outputcar1.mp4') Change to .mp4 file to test your own video. ### Go to run.py(or run2.py and find where to change by yourself)line 50-78 and 109-136 I use similar geometric relations to calculate the distance,which means you need to get the Intrinsic Matrix K by Matlab or any other code and target's height or width first.Then just change the parameters to yours. ### Go to run.py(or run2.py and find where to change by yourself)line 205 > c = cv2.waitKey(1) & 0xFF change the waitkey to get the frame played slower.
评论
    相关推荐
    • DNN FCKeditor Provider
      DNN FCKeditor Provider
    • yolov3-savevideo保存视频基于OPENCV-DNN
      yolov3-savevideo保存视频基于OPENCV-DNN,YOLOv3是流行的物体检测算法YOLO的最新变种- 你只看一次。已发布的模型可识别图像和视频中的80个不同对象,但最重要的是它具有超快速且几乎与Single Shot MultiBox(SSD)...
    • noone_video:使用DNN使视频的脸部模糊
      DNN和Haar人脸检测中找到人脸。 介绍 这项工作是使用OpenCV进行视频匿名化的人脸检测实验。 在此存储库中,所有包含的方法均来自OpenCV示例,包括 Haar检测器-经典Haar级联检测器 没有作物的DNN DNN与作物 按...
    • Dnn.Enhanced-Permissions-Provider
      该提供程序将启用DotNetNuke和DNN平台中的所有可用权限。 默认情况下,仅开箱即用配置VIEW和EDIT权限。 安装我们的提供程序将启用以下功能: 页面权限:VIEW,NAV,CONTENT,IMPORT,EXPORT,MANAGE,EDIT(重新...
    • Upscale
      高档的 介绍 这项工作源自的工作,其中包含主要思想和我更新的脚本。 您会发现我保留了模型,训练有素的权重和很多不是我做的课。 但是,与video2x一次处理一个图像相比,我改进了实现以加快升级过程,我们在这里...
    • dnn
      conda create --name dnn --file requirements.txt (requirements.txt当前不起作用,TODO) 可能需要单独conda install -c mgroth0 mlib-mgroth0更新时,请使用conda install --file requirements.txt; conda ...
    • DNN5以上的中文搜索模块
      中文搜索Provider,DNN5以上的中文搜索模块,试试下载用一下吧!
    • DnnC.CookieConsent:Dnn的CookieConsent SkinObject
      DnnC CookieConsent是一个SkinObject,是使用SilkTide的“ Cookie Consent”插件创建的,并适用于Dnn CMS平台的skinObject。 您可以在以下位置找到有关Siktide脚本的文档: ://silktide....
    • Professional.DNN7
      DNN英文学习资料。Open Source .NET CMS Platform provides thorough coverage of all aspects of DNN including the latest features, API’s, and updates. Looking to ramp up your knowledge of DNN7? If so, ...
    • 离子bb-dnn
      离子bb-dnn 激活TensorRT缓存和float16日期类型 如果要激活内部在ion-bb-dnn中使用的ONNXRuntime中的TensorRT缓存或float16,请导出以下环境变量。 export ORT_TENSORRT_ENGINE_CACHE_ENABLE=1 export ORT_TENSORRT...