ODS-Praktikum-Big-Brother

Repository for the Group Project "Big Brother" at TU Berlin
The comments, code and documentation is written in english. We only use german if there isn't any other way.

Getting Started

Understanding the project

1. Setup ssh-key.
2. Clone repository into your local folder (git clone <ssh-url>)
3. Read our contribution guidelines here!
4. Browse through our documentation

Project setup

You have a few options when executing the project locally. You can run it in the docker container or install the packages locally (in an environmen for instance) to run the flask server. Be aware that the docker container takes quite some time to build. Most of the time it would be more practical to have a python environment setup.

Docker

In this case you only need to setup docker. You may either use a docker GUI or the CLI. In case you use the CLI:

1. To into the root of the git repository.
2. Execute docker build -t bigbrother .. This takes quite a long time and requires an internet connection.
3. Execute docker run -p 3000:3000 bigbrother:latest. The 3000 refers to the port exposed in the docker container and the second 3000 is the port that you expose locally.
4. After waiting for a few seconds you should be able to go to 127.0.0.1:3000 in your browser and use the website.

Local packages

You can also do the following steps in an environment. It's important to mention again that we use Python 3.10. Other versions are not guaranteed to work, although you are welcome to try in future iterations of the project. All steps where you use the commandline are executed inside of the root directory of the git repository:

1. Install Python 3.10. Make sure that you also have the package manager pip installed. This should be the case if you installed python with the installer from the official website.
2. You might need to install the visual c++ redistributables if you are working on windows: https://learn.microsoft.com/en-us/cpp/windows/latest-supported-vc-redist?view=msvc-170
3. Install dlib. You can execute either of those:
   • pip install make cmake and then pip install dlib
   • pip install make cmake and then pip install dlib-binary
   • Build it directly from the source
4. Install the rest of the requirements with pip install -r requirements.txt.
5. Execute python ./src/bigbrother/run.py to start the flask app. You can then go to 127.0.0.1:3000 in your webbrowser.

Documentation

We have a documentation in the docs-folder. We highly recommend you to at least skim through the documetnation in order to get a grasp of the project. And the folder structure that we decided upon if you want to contribute to/extend the codebase.

Tool stack

• Programming language: Python 3.10
• Frontend: Flask Docker
• FaceRecognition and other ML-libraries:
  • openCV
  • OpenFace
  • PyTorch
  • TensorFlow
• Database: MongoDB

Benchmark

Gesture Recognition

The Benchmark loads the trained model from checkpoint and evaluate it. You can also use it to finetune the model.
Go to benchmarks/gesture_recognition and RUN: docker build -t gesture-recognition . docker run --gpus all -v $(pwd):/app -it gesture-recognition python benchmark2.py