bash terminal (although Windows / Mac OS can be used too with different commands)
gcloud (and python-app-engine using gcloud)
nvm (install node and npm using nvm)
Android SDK
adb (for installing android app in mobile device)- Create google cloud project, get its PROJECT_ID *Enable Video Intelligence API for this project
- Create a firebase project associated with this project
- In Firebase Storage Website
- Create firebase storage bucket with name STORAGE_NAME, this will contain all the videos recorded by user through the android application
- Make public in rules
service firebase.storage {
match /b/{bucket}/o {
match /{allPaths=**} {
allow read, write;
}
}
}- In Firebase Realtime Database Website
- Create a database in test mode (if creating for the first time)
- Create two firebase realtime databases with names LABELS_DATABASE_NAME and STORIES_DATABASE_NAME, these two databases will hold labels generated by video-intelligence-labelling-service and stories generated by nltk-story-generation-service respectively
- Make public in rules
{
"rules": {
".read": true,
".write": true
}
}- In Firebase Project Settings Website
- Go to service accounts
- Generate new private key
- Let the downloaded authorization file be called AUTH_FILE
- In appengine-services directory
- Copy AUTH_FILE in /video-intelligence-labelling-service directory
- Copy AUTH_FILE in /nltk-story-generation-service directory
- In /video-intelligence-labelling-service/main.py, update STORAGE_NAME, LABELS_DATABASE_NAME and AUTH_FILE variables
- In /nltk-story-generation-service/main.py, update STORIES_DATABASE_NAME and AUTH_FILE variables
- Deploy these two services as a part of the same app engine project
gcloud app deploy video-intelligence-labelling-service/app.yaml
nltk-story-generation-service/app.yaml - video-intelligence-labelling-service will generate labels from videos it accesses from the STORAGE_NAME and store it in LABELS_DATABASE_NAME, while nltk-story-generation-service will generate story for these labels amd store it in STORIES_DATABASE_NAME. For the app engine project to interact successfully with firebase project, it needs AUTH_FILE
- In firebase-functions directory
- In /functions/index.js, update variables PROJECT_NAME, STORAGE_NAME and LABELS_DATABASE_NAME
- Use node version 8
nvm use 8
- Install firebase-tools globally
npm install -g firebase-tools
- Login to to the firebase
firebase login
- Initialize firebase functions (choose N for file override, choose Y for installing npm dependencies)
firebase init functions
- Deploy firebase functions
firebase deploy --only functions
- These firebase functions, triggered by change in STORAGE_NAME and LABELS_DATABASE_NAME, make a RESTful API call to video-intelligence-labelling-service and nltk-story-generation-service respectively, passing them necessary information through query parameters
- In Firebase Authentication Website
- Go to Sign-in method
- Enable Anonymous for Sign-in providers, this is required since we are not asking users for authentication right now
- In Project Overview Website
- Click on android icon in “Add an app to get started”
- Enter android package name as “com.cloud.videostorytelling”
- Click on Register app
- Click on “Download google-services.json”, this is our config file for android project
- Click on X icon to quit out of the steps
- In android-app directory
- Copy “google-services.json” to /app directory
- In /app/src/main/java/com/cloud/videostorytelling/Camera2VideoFragment.java, update STORAGE_NAME, LABELS_DATABASE_NAME and STORIES_DATABASE_NAME.
- Run the following bash command bash gradlew clean build
- Connect Android OS based mobile device to the system through USB cable and run the following bash command
adb install ./app/build/outputs/apk/debug/app-debug.apk
- Jay Shah (@jaygshah)
- Aditya Govardhan (@AdityaGovardhan)
- Kunal Suthar (@Kunal30)