arlo-timelapse-lambda

Simpe AWS lambda function to create a timelapse movie by creating periodic snapshots using a NetGear Arlo Pro Camera and ffmpeg (on a local machine)

Introduction

When looking for a way to create a timelapse using a Arlo camera I bumped into Jeffrey d Walter's Arlo python library that has all the functionality required. Running the code on AWS lambda provides a cheap (first year should be within the AWS free tier) and hassle-free way to host such a repeating task.

The script works like this:

• AWS triggers a snapshot every 10 minutes (configurable in AWS)
• Lambda function triggers a snapshot through the Arlo cloud endpoint
• Lambda function retrieves and stores the snapshot in a private S3 bucket
• On your desktop: grab all images and construct a timelapse uing ffmpeg

Below are the steps I took (roughly) to get everything up and running on OSX (other platforms should also work, but the commands will be a bit different)

This script is open-sourced under the MIT license; please use and distribute as you wish. There are no warranties; use at your own risk. Contributions are welcome

Setup environment and dependencies

Install python - or you can install it using homebrew (OSX)

Install pipenv

pip3 install --user pipenv

depending on your install the command could be 'pip' instead of 'pip3'. I had to run the commands below to set up the enviroment properly

export LC_LOCAL=en_US.UTF-8
export LANG=en_US.UTF-8

Clone this project

git clone https://github.com/Notalifeform/arlo-timelapse-lambda

Install required pyhton libraries

In your project directory run

pipenv install

Install aws CLI

The aws library boto3 uses the aws CLI configuration to access S3, we'll also use it later to retrieve the snapshots.

brew install awscli 

AWS setup S3

First sign up for an AWS account

Make sure the aws cli and python can write to aws S3

• create bucket
• create user
• assign S3 full rights to user
• setup was cli access key locally

see https://boto3.readthedocs.io/en/latest/guide/quickstart.html

Test locally

set up your environment

export ARLO_USERNAME=user@example.com                                           
export ARLO_PASSWORD=password
export S3_BUCKET_NAME=my.bucket.name 

then run

make run 

it should output something like

pipenv run python ./arlosnapshot.py
Looking up sequence number
Creating ARLO snapshot 1
retrieving and storing  snapshot_2018-02-10_23-03-18_000000143.jpg
Script complete at 2018-02-10 23:03:24.209948

Create distribution

in your project directory run

make clean && make build

Upload your code to AWS Lambda

This is a good start to create a first version of your function

https://docs.aws.amazon.com/lambda/latest/dg/with-scheduledevents-example.html

When you upload delivery.zip make sure to set the handler to arlosnapshot.lambda_handler

The trigger you can set to something like */10 6-18 ? * MON-FRI * (in this case: weekday create a snapshot every 10 minutes from 6AM to 6PM - GMT)

Downloading pictures and creating the timelapse

Install dependencies

brew install libvpx
brew install ffmpeg --with-libvpx

Create the timelapse

Run from your snapshot dir (asuming it is next to your project dir)

../snapshot-lambda/tools/create-slideshow.sh

this should

• download all your snapshots
• symlink them so ffmpeg can find them
• create a movie using ffmpeg

the ffmpeg command line will probably need some tuning to fit your situation/preferences

Thank you

Big thanks to jeffdwalter for providing the pyhton library and quickly fixing a threading issue that I ran into using this library on AWS Lambda.

See also

• arlo pyhton library - https://github.com/jeffreydwalter/arlo
• make files for AWS Lambda - https://github.com/browniebroke/hello-lambda
• ffmpeg commands for creating slideshows - http://hamelot.io/visualization/using-ffmpeg-to-convert-a-set-of-images-into-a-video/