Galliot now hosts all Neuralet’s content and expertise gained in three years of work and completing high-quality applications, mainly in Computer Vision and Deep Learning. More Info
This repository contains the codebase related to our article on creating a face anonymizer application with Nvidia DeepStream. galliot have published three other articles which walk you through the features of Nvidia DeepStream and its python Bindings. Part 1 Part 2 Part 3
- X86 device with a Nvidia GPU
- CUDA 11.1+
- Docker and Nvidia Docker Toolkit
- Clone The repository:
git clone https://github.com/galliot-us/deepstream-example.git
cd deepstream-example
- build the docker image:
docker build -f face_annonymizer_ds_51_x86.Dockerfile -t "galliot/face_anonymizer_ds:x86" .
- Run the docker image:
docker run -it --runtime nvidia --gpus all -v "$PWD/":/repo galliot/face_anonymizer_ds:x86
- Run the
prepare_models.bash
to download the object detectors and build the required modules:
bash ./prepare_models.bash
- run one of runner scripts based on the chosen object detector:
For SSD MobileNet V2 Detector:
python3 deepstream_face_anonymizer_ssd.py --config config_infer_primary_ssd_mobilenet.txt --label_path labels.txt --input_video [PATH OF THE INPUT VIDEO] --out_dir [PATH OF THE OUTPUT VIDEO FILE]
For YOLO V3 SPP Detector:
python3 deepstream_face_anonymizer_yolo.py --config config_infer_primary_yolo_mobilenet.txt --label_path labels.txt --input_video [PATH OF THE INPUT VIDEO] --out_dir [PATH OF THE OUTPUT VIDEO FILE]