Lab course Horse detection

• Repository based on https://github.com/qqwweee/keras-yolo3, to train follow the instructions there
• Copy Folders marked in gitignore to run the examples

Usage

Run python horsinaround.py with the following arguments:

usage: horsinaround.py [-h] [--skip [SKIP]] [--single] [--multiple]
                       [--mode [MODE]]
                       input output frames

• skip should be the amount of frames you want to skip at the beginning at the video (defaults to 0).
• --single should be passed without parameter if a single horse should be tracked
• --multiple should be passed without parameter if multiple horses should be tracked
• --mode is the horse detection mode. Can be one of tinyyolo, yolo and background
• input is the input video file path
• output is the path for the video output. The .avi extension will automatically be appended, so only the filename needs to be given here
• frames is the amount of frames that should be processed. 0 to go to end of file

Dependencies

• Global
  • libopencv-dev
  • python-opencv
• Python (3.6.9)
  • tensorflow (1.6.0)
  • keras (2.1.5)
  • numpy
  • imutils
  • opencv-python
  • Pillow
  • matplotlib
  • h5py
  • cvutils
  • opencv-contrib-python (4.0.0.21)
  • progress
  • filterpy

Aufbau

cropper

• takes a frame and gives back a cropped out part in the correct aspect ratio

smoother

• takes x,y,h and gives back smoothed versions of them
• uses the last 50 frames

reader

• read
• skipframes

detector

• uses box detector
• gives back box

tracker

writer

horsinaround

• main
• argparse