Skip to content

Latest commit

 

History

History
126 lines (104 loc) · 5.05 KB

getting_started.md

File metadata and controls

126 lines (104 loc) · 5.05 KB
Raw

Getting Started with EduSense

Contents

  1. Requirements
    1. Operating Systems
    2. Dependencies
  2. EduSense Component Overview
  3. Deploying EduSense
  4. Using EduSense APIs for Client Development

Requirements

Operating Systems

We packaged all components of our system into Docker containers. EduSense should be deployable on any Linux distributions with proper docker daemon installed. Currently, we do not support Windows/Mac deployments.

Dependencies

  1. Download cuda : cuda installation. Reboot the machine, after installing cuda

  2. Check the driver :

lspci -nnk

Output should contain:

00:05.0 VGA compatible controller [0300]: NVIDIA Corporation GP102 [GeForce GTX 1080 Ti] [10de:1b06] (rev a1)
    Subsystem: eVga.com. Corp. GP102 [GeForce GTX 1080 Ti] [3842:5390]
    Kernel driver in use: nvidia
    Kernel modules: nvidiafb, nouveau, nvidia_drm, nvidia
00:06.0 Audio device [0403]: NVIDIA Corporation GP102 HDMI Audio Controller [10de:10ef] (rev a1)
    Subsystem: eVga.com. Corp. GP102 HDMI Audio Controller [3842:5390]
    Kernel driver in use: snd_hda_intel
    Kernel modules: snd_hda_intel
  1. Optional If nouveau driver grabs GPUs in step 2, follow the instructions for blacklisting nouveau.

Open /etc/modprobe.d/blacklist-nvidia-nouveau.conf and add

blacklist nouveau
options nouveau modeset=0

Then run

sudo update-initramfs -u

Reboot the machine and confirm the driver in use-:

lspci -nnk | grep "NVIDIA"

The output should look like:

00:05.0 VGA compatible controller [0300]: NVIDIA Corporation GP102 [GeForce GTX 1080 Ti] [10de:1b06] (rev a1)
    Subsystem: eVga.com. Corp. GP102 [GeForce GTX 1080 Ti] [3842:5390]
    Kernel driver in use: nvidia
    Kernel modules: nvidiafb, nouveau
00:06.0 Audio device [0403]: NVIDIA Corporation GP102 HDMI Audio Controller [10de:10ef] (rev a1)
    Subsystem: eVga.com. Corp. GP102 HDMI Audio Controller [3842:5390]
    Kernel driver in use: snd_hda_intel
    Kernel modules: snd_hda_intel

  1. Install nvidia-docker: nvidia-docker installation doc.

    • When you are done installing Docker, please make sure you also follow post-installation steps
    • Note: With Docker >= 19.03, GPUs are natively supported by Docker daemon, but docker-compose does not support GPU allocation as of now. We stick to to-be-deprecated nvidia-docker for current deployment. Once the new version of docker-compose gets released, we will explore ways to use the new functionality.

  2. Run-:

docker run --runtime=nvidia --rm nvidia/cuda nvidia-smi
  1. Install docker-compose: docker-compose installation doc

EduSense Component Overview

EduSense is composed of multiple components packaged as separate Docker components.

  • Compute Pipeline
    • openpose runs native openpose application with additional minimal API that feeds the output to video pipeline. This part is referred as Scene Parsing in the paper.
    • video post-processes and featurizes the openpose output. This component is noted as Video Featurization Modules in the paper. This module can write the output to JSON / JPG / video files or send it out to our storage backend.
    • audio processes audio data end-to-end. This module can send the output to our storage backend. This component is referred as Audio Featurization Modules in the paper.
  • Storage Backend
    • storage supports REST API end points for the compute pipelines mentioned above and EduSense client applications. Our storage backend depends on MongoDB as database backend.
  • Frontend (not included in this repository)
    • frontend visualizes processed data, providing useful debug interface. Currently, our edusense repository does not contain this component.

Deploying EduSense

To faciliate deployment, we packaged each component of EduSense into Docker containers. We provide Dockerfiles for each components and docker-compose files that automates multi-container deployment.

We provide a set of docker compose files that automates compilation and deploys EduSense. Feel free to take a look at the directory for more information.

Using EduSense APIs for Client Development

Storage server supports a set of HTTP RESTful APIs for custom EduSense applications (App Layer in the paper). Please refer to this documentation for more details about the client API.

Contributing to EduSense

We are glad to see you interested in developing with us! We always welcome new contributors. Please feel free to discuss what you found or what you want to do in our issue page.