Skip to content

How to install

Jump to bottom
Orlando Mota Pires edited this page Nov 7, 2024 · 21 revisions

Windows

Using anaconda to create an environment on Windows and install the packages

Install Miniconda (The smaller version of anaconda just with the terminal)

Open an Miniconda Terminal

  • Press the windows key at your keyboard and type 'anaconda' and them press enter.

With the terminal open, go to the conda packages installation to proceed installing all the packages and came back here to finish the process.

Download the landmark model

curl -O http://dlib.net/files/shape_predictor_68_face_landmarks.dat.bz2
bzip2 -d shape_predictor_68_face_landmarks.dat.bz2

The end! Now you are all set to work with the prototype

Running errors on windows

If there are errors while extracting features from YouTube, you'll need to update the youtube_dl and pafy packages. Just follow the instructions below:

  • Editing this file: \anaconda3\envs<bio-signals-dataset_env>\Lib\site-packages\youtube_dl\extractor\youtube.py in line 1794 as follows:

    Replace:

      1794: 'uploader_id': self._search_regex(r'/(?:channel|user)/([^/?&#]+)', owner_profile_url, 'uploader id') if owner_profile_url else None

    With:

      1794: 'uploader_id': self._search_regex(r'/(?:channel|user)/([^/?&#]+)', owner_profile_url, 'uploader id', fatal=False) if owner_profile_url else None

  • Editing this file: \anaconda3\envs<bio-signals-dataset_env>\Lib\site-packages\pafy\backend_youtube_dl.py in lines 53 and 54 as follows:

    Replace:

      53: self._likes = self._ydl_info['like_count']
  54: self._dislikes = self._ydl_info['dislike_count']

    With:

      53: self._likes = self._ydl_info.get('like_count', 0)
  54: self._dislikes = self._ydl_info.get('dislike_count', 0)

Linux

Using Miniconda to create an environment on Linux and install the packages

To install Miniconda just open your linux terminal pressing ctrl + t and run the following commands.

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
sudo chmod 777 Miniconda3-latest-Linux-x86_64.sh
./Miniconda3-latest-Linux-x86_64.sh
rm Miniconda3-latest-Linux-x86_64.sh

After installing miniconda RESTART YOUR TERMINAL

With the terminal open, go to the conda packages installation to proceed installing all the packages and came back here to finish the process.

Download the landmark model

curl -O http://dlib.net/files/shape_predictor_68_face_landmarks.dat.bz2
bzip2 -d shape_predictor_68_face_landmarks.dat.bz2

The end! Now you are all set to work with the prototype

Running errors on linux

If there are errors while extracting features from YouTube, you'll need to update the youtube_dl and pafy packages. Just follow the instructions below:

  • Editing this file: \anaconda3\envs<bio-signals-dataset_env>\Lib\site-packages\youtube_dl\extractor\youtube.py in line 1794 as follows:

    Replace:

      1794: 'uploader_id': self._search_regex(r'/(?:channel|user)/([^/?&#]+)', owner_profile_url, 'uploader id') if owner_profile_url else None

    With:

      1794: 'uploader_id': self._search_regex(r'/(?:channel|user)/([^/?&#]+)', owner_profile_url, 'uploader id', fatal=False) if owner_profile_url else None

  • Editing this file: \anaconda3\envs<bio-signals-dataset_env>\Lib\site-packages\pafy\backend_youtube_dl.py in lines 53 and 54 as follows:

    Replace:

      53: self._likes = self._ydl_info['like_count']
  54: self._dislikes = self._ydl_info['dislike_count']

    With:

      53: self._likes = self._ydl_info.get('like_count', 0)
  54: self._dislikes = self._ydl_info.get('dislike_count', 0)

Conda packages installation

Update Conda packages

conda update --all

Create a conda environment (Here an environment called "bio-signals-dataset_env" will be used to install the packages)

conda create --name bio-signals-dataset_env python=3.10 pip cmake ipython jupyter

Activate the environment

conda activate bio-signals-dataset_env

Install conda packages

conda install -c conda-forge dlib opencv numpy pandas matplotlib imutils stumpy plotly

Install pip packages

pip install youtube_dl==2021.12.17 pafy==0.5.5 moviepy==1.0.3 notebook dash scikit-learn imbalanced-learn optuna tslearn seaborn

Using tensorflow for CPU

If you want to use the tensorflow running on a GPU (recommended), go here for LINUX and here for WINDOWS, otherwise, if you want to use on the CPU (not recommended) run the following command.

pip install tensorflow

Using Tensorflow for GPU on Linux

If you want to use tensorflow on your GPU for training and making inferences at the neural net, just follow the installation guide here bellow.

  1. Install the latest NVIDIA driver version via GeforceExperience here

  2. Install the latest cuda tool-kit version here

After installing the CUDA tool-kit, run the tensorflow gpu installation for conda (make sure you are at the bio-signals-dataset_env environment previously created)

conda install -c conda-forge tensorflow-gpu

Using Tensorflow for GPU on Windows

If you want to use your GPU for inferences or training the neural network follow the steps bellow

  1. Install the latest NVIDIA driver version via GeforceExperience here

  2. Install the cuda-toolkit version 11.8 here

  3. Install tensorflow via pip (THE TENSORFLOW VERSION MUST BE BELOW 2.11) and cudnn via conda-forge

     pip install "tensorflow<2.11"
 conda install -c conda-forge cudnn=8.1

If there is any doubts left, go at the complete guide here

Clone this wiki locally