USAGE.md

Scripts

1. Get my OAuth2 IDs

2. Get a json key file and put it in key_account directory

3. Fill configure_gcp.sh exports with your gcp config

chmod +x configure_gcp.sh
./configure_gcp.sh

then follow following instructions (or check configure_gcp_and_deploy.sh but not tested)

gcloud CLI

echo "deb [signed-by=/usr/share/keyrings/cloud.google.gpg] https://packages.cloud.google.com/apt cloud-sdk main" | sudo tee -a /etc/apt/sources.list.d/google-cloud-sdk.list \
&& curl https://packages.cloud.google.com/apt/doc/apt-key.gpg | sudo apt-key --keyring /usr/share/keyrings/cloud.google.gpg add - \
&& sudo apt-get update \
&& sudo apt-get install -y google-cloud-sdk \
&& sudo rm -rf /var/lib/apt/lists/*

Create GCP storage bucket

gsutil mb gs://$BUCKET_NAME/ \
--regions $REGION
gsutil defacl set public-read gs://$BUCKET_NAME

Deploy an object detection model to AI Platform

1. Pick a model from tensorflow models or somewhere else (AI Platform deployable format: SavedModel)

2. To deploy a model either

   • use Cloud Run service (recommended) (see docs)

   Or use

   • use Google Colab

3. Then push class mapping to datastore

Deploy Cloud Function

Input Pub/Sub

gcloud functions deploy queue_input \
--source cloud_functions/queue_input \
--runtime python37 \
--project $PROJECT_ID \
--trigger-http \
--region $REGION \
--env-vars-file cloud_functions/queue_input/.env.yaml \
--memory 2gb

Predictor

gcloud functions deploy predictor \
--source cloud_functions/predictor \
--runtime python37 \
--project $PROJECT_ID \
--trigger-http \
--region $REGION \
--env-vars-file cloud_functions/predictor/.env.yaml \
--max-instances 1 \
--memory 2gb

Batch result

gcloud functions deploy batch_result \
--source cloud_functions/batch_result \
--runtime python37 \
--project $PROJECT_ID \
--trigger-resource gs://$BUCKET_NAME \
--region $REGION \
--trigger-event google.storage.object.finalize \
--env-vars-file cloud_functions/batch_result/.env.yaml \
--max-instances 1 \
--memory 2gb

Frame extractor

gcloud functions deploy extractPubSub \
--source cloud_functions/frame_extractor \
--runtime nodejs10 \
--project $PROJECT_ID \
--trigger-topic topic_extractor \
--region $REGION \
--env-vars-file cloud_functions/frame_extractor/.env.yaml \
--max-instances 1 \
--memory 2gb

Dont take all my money

follow to avoid having your bank account emptied by Google

THIS HASNT BEEN TESTED YET SO FOLLOW THE LINK INSTRUCTIONS CAREFULLY

gcloud functions deploy stop_billing \
--source cloud_functions/dont_take_all_my_money \
--runtime python37 \
--project $PROJECT_ID \
--trigger-topic budget-notifications \
--region $REGION \
--env-vars-file cloud_functions/dont_take_all_my_money/.env.yaml

Or

gcloud functions deploy limit_use \
--source cloud_functions/dont_take_all_my_money \
--runtime python37 \
--project $PROJECT_ID \
--trigger-topic budget-notifications \
--region $REGION \
--env-vars-file cloud_functions/dont_take_all_my_money/.env.yaml

Deploy to Google Cloud App engine

gcloud config set project $PROJECT_ID
gcloud app deploy
gcloud scheduler jobs create http cron_input --schedule='every 30 mins' --uri="https://$REGION-$PROJECT_ID.cloudfunctions.net/queue_input"