-
This repository (including the submodules) was manually copied from bendidi/Tracking-with-darkflow.
- The original
README.mdhas been moved toREADME_orig.md - Copy date: 2017/12/19
- The original
-
I tried this repository in Ubuntu 16.04 within a conda environment (FYI:
repo/conda_env.yml).
In 2017, the OpenCV installed by conda had problems with video reading/writing functions.
In 2018 (at the posting time, it was 2018-01-23), just install OpenCV from conda-forge solves the problem:
conda install -c conda-forge opencv
-
Make sure to install OpenCV manually before installing Anaconda.
- I had tried to build OpenCV when my system already has Anaconda 3 installed, and it never worked.
- This step is just a workaround. If you know better approaches, welcome to leave me a message.
-
After build and install OpenCV in your system, do the following things to install Anaconda and make it access to your system's OpenCV:
-
install Anaconda 3
-
create a test environment by
conda create -n <your_env_name> python=3.6(or userepo/conda_environment.yml) -
go to
~/.conda/envs/<your_env_name>/lib/python3.6/site-packages -
link the so file by
ln -s /usr/local/lib/python3.5/dist-packages/cv2.cpython-35m-x86_64-linux-gnu.so cv2.so -
to check whether it works, you could go to
repoand runpython check_vid_play.py <test_video_file>.
-
- Download the repository, go to
darkflowfolder, then runpython3 setup.py build_ext --inplace. - Edit the
run.pyto fit your mission. - Run
python run.pyand watch the results. :-)