Social_distancing_with_Detectron2

Modules needed:

pip install -U torch torchvision cython
pip install -U 'git+https://github.com/facebookresearch/fvcore.git' 'git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI'
git clone https://github.com/facebookresearch/detectron2 detectron2_repo
pip install -e detectron2_repo
pip install scipy
pip install numpy

You will also need opencv Make sure your python version is 1.4+

How does it work

I used detectron2 to get the predictions and used my own visualizing module (image_viz.py) to bound-box only humans. The later part of the code which deals with social distance was completed with the help of this article which I found way later after starting this project and realised that they also have used Detectron2 for their detection (If I had found this article slightly earlier, it would have saved a lot of time). The main inspiration for this article came from this medium post

How to run the code

First, it is highly suggestible to download the model and keep it in the main folder.

python runner.py --config-file <path_detectron2_repo>/configs/COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml \
--input cctv-footage.jpg  --opts MODEL.DEVICE cpu  MODEL.WEIGHTS model_final_f10217.pkl 

• To run on your webcam, replace --input files with --webcam.
• To run on a video, replace --input files with --video-input video.mp4.
• To run on cpu, add MODEL.DEVICE cpu after --opts.
• To save outputs to a directory (for images) or a file (for webcam or video), use --output.