Skip to content

Latest commit

 

History

History
69 lines (37 loc) · 2.05 KB

README.md

File metadata and controls

69 lines (37 loc) · 2.05 KB
Raw

catfinder5000-parse

Overview

Overview of “Parsing” lambda

CloudWatch triggers a parsing of the video URL every minute. Get a time reference from Program-Date-Time information Create a static JPG image from the scene in the video segment, stored in Amazon S3 Invoke Amazon Rekognition to tell what is on the scene, with what confidence level Store the Time reference information, frame numbers and Delta URL in the Amazon DynamoDB

Why I designed it this way

We are CURL-ing the file to local file system via a loop in a Lambda Function that CloudWatch Events is triggered every 1 minute. This design choice was to keep this lab condensed into the time allotted. It is not the most effecient way, just the easiest to keep this lab under 2.5 hours. A interesting side effect is this method could be used on a livestream on any origin/CDN. Other methods such as pushing HLS directly to a S3 bucket for S3 Event Triggers for Lambda Functions would require custom configurations for the Elemental Live encoder. Using this method helps keep our Lab within the correct span of time.

Code Steps

  1. download m3u8s save to /tmp

  2. parse m3u8s to get stream and time info

  3. download .ts video file

  4. ffprobe for PTS values and scene change analysis

  5. ffmpeg for thumbnails of every frame to /tmp

  6. identify each thumbnail for a scenechange indication

  7. upload thumnail to S3 bucket

  8. Rekognition anaylsis of thumbnail

  9. if "person" label, use Celebrity Rekognition

  10. put all info into DynamoDB

  11. if "Cat" label detected invoke "prekog" Lambda Function

  12. delete jpg and ts files

  13. write json file from DynamoDB info

Environment Variables

Mandatory

S3_BUCKET

HLS_URL

Optional

you can override the defaults listed below with the Enviornment Variable.

DYNAMO_MAIN = "catfinder5000-main"

DYNAMO_LIST = "catfinder5000-list"

DYNAMO_SUMMARY = "catfinder5000-summary"

DYNAMO_SUMMARY_GSI = 'rekog_type-timestamp_updated-index'

LAMBDA_PREKOG = "catfinder5000-prekog"

REKOG_LABEL = "Cat"