Video Intelligence API keynote demo from Google Next 2017
Clone or download
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
backend Update to use v1 Video API Dec 3, 2017
frontend Update README and keyfile path in app.js Dec 4, 2017
sample-annotations Update to use v1 Video API Dec 3, 2017
.gitignore Update gitignore Jun 8, 2017
CONTRIBUTING.md Initial commit Apr 13, 2017
LICENSE Initial commit Apr 13, 2017
README.md Update README and keyfile path in app.js Dec 4, 2017
architecture.png Initial commit Apr 13, 2017

README.md

This is not an official Google product.

Video Intelligence API Demo

This is the code for the Video Intelligence API demo presented at Google Cloud Next 2017. See a video of the presentation here. Big thank you to Alex Wolfe for his contributions to this app.

The code for the app is split into frontend and backend repos. Here's what it looks like:

Architecture diagram

FRONTEND

  • App Engine app that displays videos and their Video API annotations, and lets you search videos by annotation.

TECH STACK

BACKEND

  • Google Cloud Function that calls the Video API everytime a new video is added to a bucket.
  • It stores the JSON response output a separate GCS bucket.

TECH STACK

RUNNING THE APP

Prerequisites

Setting up the backend (Cloud Functions + Video Intelligence API)

Setup Project

  1. Create a Cloud project
  2. Enable the Video Intelligence API
  3. Enable Cloud Functions
  4. Generate a JSON keyfile

Setup Storage

  1. In your project, create four Cloud Storage buckets:
    • one for your videos
    • one for the video JSON output
    • one for the video thumbnails
    • one as a staging bucket for your Cloud Function
  2. Set permissions for your Storage Buckets
    • Make video bucket world readable - Group - allUsers - Reader
    • Make JSON annotation bucket writable by service account - User - [serviceaccount] - Owner

Setup Credentials

  1. Make copy of frontend/local.sample.json named frontend/local.json
  2. Make copy of backend/local.sample.json named backend/local.json
  3. Copy your Cloud project ID and storage bucket names into frontend/local.json and backend/local.json
  4. Copy your generated service account json file into a file called keyfile.json and place a copy in both your frontend AND backend directories (you'll deploy these separately, one to App Engine and one to Cloud Functions)

Setting up the local development copy of the frontend

  1. Clone this repo
  2. cd into the frontend directory
  3. Run npm install to install dependencies
  4. Run npm start in one tab on your terminal
  5. Create a new tab and gulp dev init
  6. Make sure npm start and gulp are running at the same time
  7. Navigate to localhost:8080. You should see the UI without any videos - that part is next!
  8. (Optional step) If you want to see the UI with some sample video content before deploying your function and adding your own videos, copy the files from the sample-annotations directory into your storage buckets:
    • Copygoogle-home-superbowl.mp4 into your video storage bucket
    • Copy google-home-superbowlmp4.json file to your video JSON annotation storage bucket
    • Copy both png files into your thumbnail bucket
    • Run the frontend and you'll see the video with the annotations visualized

DEPLOY

Deploy the backend

  1. cd into backend
  2. run gcloud config set project your-project-id
  3. Deploy your Cloud Functions:
    • Deploy the function for annotating videos with the Video Intelligence API, replacing the bucket names with the ones you created: gcloud beta functions deploy analyzeVideo --stage-bucket your-stage-bucket-name --trigger-bucket your-video-bucket
    • Deploy the function for generating thumbnails (these are the images you see on the homepage): gcloud beta functions deploy generateThumbnail --stage-bucket your-stage-bucket-name --trigger-bucket your-video-bucket. Note: this and the next function require you to deploy the ffmpeg binaries with your functions code. You can download them here.
    • Deploy the function for generating previews (these are the images you see on an individual video page): gcloud beta functions deploy generatePreview --stage-bucket your-stage-bucket-name --trigger-bucket your-video-bucket
  4. With your functions deployed, try uploading a video to your video storage bucket. When the Video API finishes processing, you should see the annotation JSON file in your annotation bucket and your thumbnail images in your thumbnail bucket. If the JSON annotations or thumbnails aren't there, check your Functions logs for errors in the Logging tab of your Cloud console
  5. Note: you may need to increase the time and storage limits on your Cloud Functions. Do this by navigating to the Cloud Functions section of your Cloud console, selecting the function, then clicking "Edit". Next update "Memory allocated" in the dropdown andselect "Configure region, timeout and retry policy" to extend the timeout.
  6. To see the video in your UI: navigate to localhost:8080/profile, then click 'clear indexeddb' and 'get videos'. Whenever you add new videos you'll need to clear the local cache and re-download to see them.

Deploy the frontend

  1. cd into frontend
  2. run gcloud config set project your-project-id
  3. Deploy your app with gcloud app deploy