Prerequisites

• Disable constraints/cloudfunctions.allowedIngressSettings org policy on the project
• Disable constraints/iam.disableServiceAccountKeyCreation org policy on the project
• Enable all the APIs https://console.cloud.google.com/flows/enableapi?apiid=cloudfunctions,cloudbuild.googleapis.com,pubsub,storage_api,vision.googleapis.com&redirect=https://cloud.google.com/functions/docs/tutorials/ocr&_ga=2.154163455.673635666.1645538540-966849366.1644425268
• Give default app engine svc acct owner privs (fix later but needs at least data editor for BQ insert)
• Used this as a guide: https://cloud.google.com/functions/docs/tutorials/ocr#functions_ocr_setup-python
• In BQ, create:
  • dataset named "routes"
  • table named "test-data"
  • create a table called test-data and upload table_schema.json and select auto-detect schema

Setup

0. Create the bucket that will store the route images, the bucket that will receive the route metadata CSV and trigger the cloud function, and the bucket where the resulting JSON file containing the items found in the route pictures and their location will be stored

   gsutil mb gs://route_images_02
   gsutil mb gs://route_metadata_02
   gsutil mb gs://route_results_02
   

1. Upload images to gsutil mb gs://route_images_02

   gsutil -m cp ./sample_data/*.jpg gs://route_images_02
   

2. Deploy the image processing function with a Cloud Storage trigger

gcloud functions deploy svai-extract \
--runtime python39 \
--set-env-vars RESULTS_BUCKET=route_results_02 \
--trigger-bucket route_metadata_02 \
--entry-point process_image

0. Metadata is typically generated as a single JSON file. We need to split that into one JSON file per image because event-based cloud function triggers timeout after 10 minutes. Do the following to properly split the files:
   • Copy the JSON that looks like the following in the image_metadata.json accompanying the images to metadata_splitter.sh

     "notificationId": "some_value",
     "notificationTimestamp": "some_value",
     "siteId": "some_value",
     "siteOwner": "some_value",
     "siteName": "some_value",
     

   • In metadata_splitter.sh change the filepath on the line with the jq statement to point to image_metadata.json
   • In image_metadata.json find and replace all bucket names and file paths with route_images_02
   • Run the script: ./metadata_splitter.sh
1. While the SVAI API is in alpha give it about 1 minute in between uploads to avoid 500 errors

for FILE in ./SOME_DATE/*.json; do gsutil cp $FILE gs://route_metadata_02; sleep 60; done

Cleanup

gcloud functions delete svai-extract

Demo Flow

0. gsutil cp ./sample_data/61f53b8cc6f5379721f5c7b24c100d8f8f7d9c323ef7366ff4f5d73fa2b3530331b6b1da2ec4b9054862a3808edc4bef782041d207d24b448dff7540fd158f7f.json gs://route_metadata_02
1. Cloud Function Logs
2. BigQuery Table using the following query:

   SELECT *
   FROM `cloud-store-vision-test.routes.test-data` AS testdata
   LEFT JOIN UNNEST(testdata.data) as testdata__data
   WHERE testdata__data.hash = FROM_BASE64 ('003c0cf434790ee7aabee1672e8b467c1cfe75d1d680e0c7161371f3ad67492e476c175be00f8a8107365c1aeb32621d8e8c2b92568e1ec073f7e15b8c7d1a35')