Smart Album

Smart Album is built on Google Cloud Platform, a service that can classify images in a given google photo account. After processing images, there will be a home page where user can query class name to find images slideshow of that class.

Prerequisites:

1. Google Cloud Platform
2. Google Photo API
3. Kubernetes

Steps:

1. Create GCP Computing Cluster: > ./create_cluster.sh
2. Create Image Worker base image: > ./worker/create-base-images.sh
3. Create Image worker Deployment, Service, Horizontal Scaling: > ./worker/create_worker.sh
4. Create Redis Deployment, Service: > ./redis/redis-launch.sh
5. Create Frontend Deployment, Service: > ./frontend/frontend-launch.sh
6. Set Google Photo API: Replace Authorized JavaScript origins and Authorized redirect URIs with the external ip of your Frontend Service (Run > kubectl get services to find it) After setting up Google Photo API, click Download JSON and store secret file under /frontend/auth repository with the filename client_secret.json
7. Remove existing Frontend Deployment> kubectl delete deployment rest-server and Rerun step.5 command. (An error message saying the service is already created is expected. But we can totally ignore it!)

Run:

Enter the external ip address plus 'xip.io:5000' Ex: http://34.82.36.165.xip.io:5000 The reason we requires xip.io is that we did not set DNS name for our service, yet Google Photo API requires a valid URI to work. Port 5000 is the default port used by FLASK.

Demo Video Link:

List of Software and Hardware Components:

1. Python
2. Keras (image processing model)
3. Flask REST server (/, /index, /authorize, /4. oauth2callback, /process_images)
4. gRPC
5. Redis (store the metadata of the JPEG)
6. Google Photos API
7. OAuth 2.0

Architecture:

Resources:

• Keras pretrained VGG16 w/ imagenet
• W3.css Frontend Template