Object Detection

• I recorded a video while playing GTA5 and tried to implement object detection on the video using YOLO-V2.

The raw video file recorded from gta5

Requirements :

• python 3.7
• tensorflow
• opencv
• numpy

Download the darkflow Repo from here

• the code written in the darkflow repo is in tensorflow version1.
• we can also migrate our tensorflow1 code to tensorflow2 using:

  import tensorflow.compat.v1 as tf
  tf.disable_v2_behavior()

Build the Library

• open an cmd window and type:

  - python setup.py build_ext --inplace

  or

  - pip install -e

Download weights file from here

Demo conversion

• move the video file into the darkflow-master
• from there, open a cmd window
• use the command:

  python flow --model cfg/yolo.cfg --load bin/yolov2.weights --demo videofile.mp4 --gpu 1.0 --saveVideo

• videofile.mp4 is the name of your video.
• NOTE: if you do not have the GPU version of tensorflow, leave off the —gpu 1.0
• —saveVideo indicates to save a name video file, which has the boxes around objects

I suggest the conversion via python using opencv

• the code is written in main.py.
• if using cpu version we can remove ‘gpu’: 1.0 from options

The processed video file