/IW276SS20-P2

Affective Computing - Face Expression Recognition

Primary LanguagePythonMIT LicenseMIT

IW276SS20P2: Affective Computing: Face Expression Recognition (Jetson Nano)

Short introduction to project assigment.


Link to Demo Video

This work was done by Daahir Ali, Louis Tschall, Konstantin Krause during the IWI276 Autonome Systeme Labor at the Karlsruhe University of Applied Sciences (Hochschule Karlsruhe - Technik und Wirtschaft) in SS 2020.

Table of Contents

Requirements

  • Python 3.6 (or above)
  • OpenCV 4.0 (or above)
  • Jetson Nano
  • Jetpack 4.2

Prerequisites

  1. Install requirements: Installation in virtual environment:
    • Pytorch 
      wget https://nvidia.box.com/shared/static/3ibazbiwtkl181n95n9em3wtrca7tdzp.whl -O torch-1.5.0-cp36-cp36m-linux_aarch64.whl
sudo apt-get install python3-pip libopenblas-base
pip install Cython
pip install numpy torch-1.4.0-cp36-cp36m-linux_aarch64.whl
    • Torchvision 
      sudo apt-get install libjpeg-dev zlib1g-dev
git clone --branch v0.6.0 https://github.com/pytorch/vision torchvision
cd torchvision
sudo python setup.py install
cd ../
    • torch2trt Follow manual on https://github.com/NVIDIA-AI-IOT/torch2trt#setup
    • Face_Recognition (https://github.com/ageitgey/face_recognition) 
      pip install face_recognition
  2. Convert pretrained model
  • Conversion 
    python convert2trt.py <mode> <source> <target>
  • Info: the pipeline expects the converted model to be located under './models/resnet50.224.trt.pth'. If needed this has to be changed in face_expression_recognition.py.

Pre-trained models

Pre-trained model is available at pretrained-models. The converteed pretrained model has to be unzipped using 7-Zip, because of GitHub's 100mb upload limit This can be done with 7z x archive.7z.001 The converted pretrained model resnet50.224.trt.pth may or may not work on your machine bunt in any case a self-converted model is signtificantly faster than this one.

Running

The pipeline can be run with

python pipeline_main.py <source>

Available sources are for example:

  • Camera-Index (e.g. 0)
  • IP-Camera (als URL)
  • Video (PfPathad, e.g. mp4)

In file config.yml are settings to improve performance or accuracy.

Acknowledgments

This repo is based on

Thanks to the original authors for their work!

Contact

Please email mickael.cormier AT iosb.fraunhofer.de or tschalllouis AT-Sign gmail.com for further questions.