Linux / amd64
This container provides a demonstration of how to deploy pre-trained models from NGC into an intelligent video analytics (IVA) pipeline in DeepStream. These models could be fine-tuned on additional data using Transfer Learning Toolkit.
NVIDIA DeepStream lets you deploy deep learning models in intelligent video analytics (IVA) pipelines. This enables real time object detection, tracking, and classification. This is possible through a combination of tools such as TensorRT, Triton, Transfer Learning Toolkit the NVIDIA Codec SDK and the NVIDIA Optical Flow SDK.
Please review the following resources:
Getting started with the application is pretty straightforward with nvidia-docker.
This image contains the complete DeepStream Jupyter Lab environment and tutorial.
1. Download the container from NGC
docker pull nvcr.io/nvidia/ds_deployment_demo:20.06
2. Run the notebook server
docker run --gpus all --rm -it -p 8888:8888 --env NVIDIA_VISIBLE_DEVICES=0 nvcr.io/nvidia/ds_deployment_demo:20.06
Note: Depending on your docker version you may have to use ‘docker run --runtime=nvidia’ or remove ‘--gpus all’
3. Connect to notebook server
Jupyter Lab will be available on port 8888!
e.g. http://127.0.0.1:8888 if running on a local machine
(or first available port after that, 8889, 8890 etc if 8888 is occupied - see command output)
4. Run the notebooks
deepstream_deployment: Introduction to DeepStream and step-by-step deployment of 4 different pre-trained models from NGC.
TLT_Sample_Model: This notebook will not run inside this container, it is merely intended to demonstrate the process is for training models in Transfer Learning Toolkit.