Skip to content

Latest commit

 

History

History
33 lines (21 loc) · 1.28 KB

README.md

File metadata and controls

33 lines (21 loc) · 1.28 KB
Raw

Human Tracking and Recognition Program.

Human detection tracking and recognition program via camera or video using Deep SORT, YOLOv3, and PCB.

Dependencies

  • Python > 3.6
  • Pytorch > 0.3
  • Tensorflow > 1.9.0

Need GPU to run smoothly.

Basic Use

  1. Train a Features Extraction model from 'Person_reID_baseline_pytorch-master' folder train PCB for basically.
    You may need to download Market1501 Dataset for training

    More detail in README.md

  2. Get pre-train YOLO model from : yolo.h5 and put it into Model folder.

  3. Use ExtractFromVid.py to crop person image from video to Sample/ALL folder. Then make subfolder and name it for each person you want program to recognize. (for other person you don't want put them into one Unknow folder)
    You may need some Pedestrian video in Video folder for do a sample.

  4. Use MakeSampleSet.py to make a sample set file and SVM model from all subfolder in Sample.

  5. Check path and parameter before run Main.py file.

Reference