Food Classifier App with TensorFlow 2 model on Google Cloud Run

This repo contains a web app for classifying 3 types of local Singaporean Food using the following APIs and platforms

• TensorFlow 2: using tf.image, tf.keras.models and tf.data (for training)
• Google Cloud Platform: Cloud Run, Cloud Build, Cloud Storage
• Starlette, a lightweight ASGI framework/toolkit

---

Deploy with Google Cloud Run

Pre-requisites

• Google Cloud Account
• Enable Cloud Run and Cloud Build API

Deploy

To run the webapp on Google Cloud Run, launch Cloud Shell from your Dashboard.

• export PROJECT_ID={your project id here}
• Run gcloud config set project $PROJECT_ID
• Run git clone https://github.com/yoke2/food_classifier_tf2_cloud_run.git
• Change directory: cd food_classififer_tf2_cloud_run
• Build container image with Google Cloud Build by running gcloud builds submit --tag gcr.io/$PROJECT_ID/food_classifier
  • If prompted that Cloud Build API not enabled on the project, enter 'y' to enable and retry
• Deploy container built: gcloud run deploy --image gcr.io/$PROJECT_ID/food_classifier --platform managed --allow-unauthenticated --memory 1024M
  • You will be prompted for the service name: press Enter to accept the default name
  • You will be prompted for region: select the region us-central1
• You will receive an URL to access the app. This can be seen on the Cloud Run page for dimsumapp in the Dashboard as well

---

Docker Installation

You can test your changes locally by installing Docker and using the following command:

docker build -t food-tf2-cr . && docker run --rm -it -p 8080:8080 food-tf2-cr

---

Local Installation

• Install dependencies

$ pip install packaging
$ pip install -r requirements.txt

Run app

$ python app/server.py serve

---

Reference

• Based on fast.ai's guide for production deployment to Render
• Google Cloud Run Docs
• Repository for supplementary package containing supporting functions