Label images by using an AutoML Vision Edge model

Label images by using an AutoML Vision Edge model

This quickstart walks you through the process of:

• Copying a set of images into Google Cloud Storage.
• Creating a CSV listing the images and their labels.
• Using AutoML Vision to create your dataset, train a custom AutoML Vision Edge model (image classification or object detection), and make a prediction.
• Exporting and deploying your AutoML Vision Edge model for one of several types of edge devices, such as mobile phones, ARM-based devices, and the Coral Edge TPU.

Caution: If you're targeting the Coral Edge TPU, beware that the current AutoML object detection model runs most of its graph ops on the CPU instead of the Edge TPU. For better performance, instead train your model using one of Coral's model retraining Colabs.

Before you begin

Region requirement: Currently project resources must be in the us-central1 region to use this API.

Set up your project

1. Sign in to your Google Cloud account. If you're new to Google Cloud, create an account to evaluate how our products perform in real-world scenarios. New customers also get $300 in free credits to run, test, and deploy workloads.
2. Install the Google Cloud CLI.

3. To initialize the gcloud CLI, run the following command:

   gcloud init

4. Create or select a Google Cloud project.

   Note: If you don't plan to keep the resources that you create in this procedure, create a project instead of selecting an existing project. After you finish these steps, you can delete the project, removing all resources associated with the project.

   • Create a Google Cloud project:

     gcloud projects create PROJECT_ID

   • Select the Google Cloud project that you created:

     gcloud config set project PROJECT_ID

5. Make sure that billing is enabled for your Google Cloud project.

6. Enable the AutoML and Cloud Storage APIs:

   gcloud services enable storage-component.googleapis.com automl.googleapis.com storage-api.googleapis.com

7. Install the Google Cloud CLI.

8. To initialize the gcloud CLI, run the following command:

   gcloud init

9. Create or select a Google Cloud project.

   Note: If you don't plan to keep the resources that you create in this procedure, create a project instead of selecting an existing project. After you finish these steps, you can delete the project, removing all resources associated with the project.

   • Create a Google Cloud project:

     gcloud projects create PROJECT_ID

   • Select the Google Cloud project that you created:

     gcloud config set project PROJECT_ID

10. Make sure that billing is enabled for your Google Cloud project.

11. Enable the AutoML and Cloud Storage APIs:

    gcloud services enable storage-component.googleapis.com automl.googleapis.com storage-api.googleapis.com

12. Set the PROJECT_ID environment variable to your Project ID.

    export PROJECT_ID=PROJECT_ID

    The AutoML API calls and resource names include your Project ID in them. The PROJECT_ID environment variable provides a convenient way to specify the ID.

Create a Cloud Storage bucket

Use Cloud Shell, a browser-based Linux command line connected to your Google Cloud console project, to create your Cloud Storage bucket:

1. Open Cloud Shell.

2. Create a Google Cloud Storage bucket. The bucket name must be in the format: project-id-vcm. The following command creates a storage bucket in the us-central1 region named project-id-vcm. For a complete list of available regions, see the Bucket Locations page.

   gsutil mb -p ${PROJECT_ID} -c regional -l us-central1 gs://${PROJECT_ID}-vcm/

3. Set the BUCKET variable.

   export BUCKET=${PROJECT_ID}-vcm

Copy the sample images into your bucket

Next, copy the flower dataset used in this Tensorflow blog post. The images are stored in a public Cloud Storage bucket, so you can copy them directly from there to your own bucket.

1. In your Cloud Shell session, enter:

   gsutil -m cp -R gs://cloud-samples-data/ai-platform/flowers/ gs://${BUCKET}/img/
   

   The file copying takes about 20 minutes to complete.

Create the CSV file

The sample dataset contains a CSV file with all of the image locations and the labels for each image. You'll use that to create your own CSV file:

1. Update the CSV file to point to the files in your own bucket:

   gsutil cat gs://${BUCKET}/img/flowers/all_data.csv | sed "s:cloud-ml-data/img/flower_photos/:${BUCKET}/img/flowers/:" > all_data.csv
   

2. Copy the CSV file into your bucket:

   gsutil cp all_data.csv gs://${BUCKET}/csv/
   

Create your dataset

Visit the AutoML Vision UI to begin the process of creating your dataset and training your model.

When prompted, make sure to select the project that you used for your Cloud Storage bucket.

1. From the AutoML Vision page, click New Dataset:

   

2. Specify a name for this dataset (optional), select the model objective, and then click Create dataset to continue the dataset creation process.

   

3. In the Select files to import screen choose the Select a CSV file on Cloud Storage radio option. Specify the Cloud Storage URI of your CSV file. For this quickstart, the CSV file is at:

   • gs://${PROJECT_ID}-vcm/csv/all_data.csv

   Replace PROJECT_ID with your specific project ID.

   

4. Click Continue to begin image import. The import process takes a few minutes. When it completes, you are taken to the next page which has details on all of the images identified for your dataset, both labeled and unlabeled images.

   When using the indicated flower dataset, you will see several warning alerts. This is purposeful, to show you error messages you may encounter with your own data.

   

Train your model

1. Once your dataset has been created and processed, select the Train tab to initiate model training.

   

2. Select Start training to continue. This will open a Train new model window with training options.

3. In the Define your model section of the new model training window, change the model name (optional) and select the Edge model radio option. Select Continue to move to the following section.

   

4. In the Optimize model for section accept the Best trade-off option and select Continue.

5. In the Set a node hour budget section accept the suggested node budget (4 node hours).

   Node hour budget: Recommended training time is calculated based on:

   • model learning curves
   • training dataset size

   If model training converges before recommended or custom time selected the system allows for early stopping; this means you are only charged for the time it takes to train the model. You are encouraged to use the recommended amount of hours to avoid using more resources than necessary and reduce billing costs.

   

6. Select Start training to begin model training.

   Training is initiated for your model, and should take about an hour. The training might stop earlier than the node hour you selected. The service will email you once training has completed, or if any errors occur.

Note: After training completes, you can refer to evaluation metrics, as well as test and use the model. Select the Evaluate tab to get more details on F1, Precision, and Recall scores. Select an individual label under Filter labels to get details on true positives, false negatives and false positives.

Deploy the model

Before you can export your model you must deploy it for use.

1. To deploy your model, select the Test & use tab. In the tab click the Deploy model option near the model name.

2. In the window that follows, specify 1 node to deploy on, and select Deploy to begin the model deployment process.

   

You will receive a notification when model deployment completes.

Export the model

The final step in using an AutoML Vision Edge model is to export (optimize and download) and deploy (use) your model.

There are multiple ways you can export and deploy your models to use for prediction on Edge devices.

Model and export options: Tensorflow Lite is optimized for mobile phones and ARM (mobile device operating system) devices. It can further be compiled to run on Edge TPU. Another export option - Tensorflow SavedModel - can be used in a Docker container for serving. Lastly, a Core ML model is specially optimized for iOS apps.

In this quickstart you will use Tensorflow Lite (TF Lite) as an example. TF Lite models are both easy to use and have a wide set of use cases.

1. Under the Use your model section of the Test & use tab, select the TF Lite option.

   

   Model export requirement: You must export an Edge model to the same Cloud Storage bucket that you used for data import.

2. In the Export TF Lite package window that shows, specify a Cloud Storage bucket location to export a TF Lite package to, then select Export. The export process typically takes several minutes.

   

In the Google Cloud Storage destination location you will find a folder named with timestamp and model format, under which you can find the following files:

• a tflite file (model.tflite),
• a dictionary file (dict.txt)
• a metadata file (tflite_metadata.json)

Clean up

To avoid incurring charges to your Google Cloud account for the resources used on this page, delete the Google Cloud project with the resources.

What's next

With these files, you can follow tutorials to deploy on Android devices, iOS devices, Raspberry Pi 3, or the Web.

Other model use options

• You can export the model as a CoreML (iOS/macOS) supported model. After training, you can export the model by selecting the CoreML option in the Test & use tab, and follow the CoreML tutorial.
• You can export the model for running on the Coral Edge TPU. After training, export the model by selecting the Coral option in the Test & use tab. However, due to current incompatibility with our object detection model, we recommend you only export image classification models for the Edge TPU (to train an object detection model for the Edge TPU, instead follow one of these Google Colab tutorials). After exporting your model, follow Coral's official documentation about how to run an inference on the Edge TPU.
• You can export the model as TensorFlow SavedModel and use it with a Docker container. After training, you can export the model by selecting the Container option in the Test & use tab, and follow the Edge containers tutorial on how to export to a container.
• You can export the model for use in a browser or in Node.js as a Tensorflow.js model. After training, you can export the model by selecting the Tensorflow.js option in the Test & use tab, and follow the Edge TensorFlow.js tutorial.

Cleanup

If you no longer need your custom model or dataset, you can delete them.

To avoid unnecessary Google Cloud Platform charges, use the GCP Console to delete your project if you do not need it.

Undeploy your model

Your model incurs charges while it is deployed.

1. Select the Test & Use tab just below the title bar.
2. Select Remove deployment from the banner beneath your model name to open the undeploy option window.

   

3. Select Remove deployment to undeploy the model.

   

4. You will receive an email when model undeployment has completed.

Delete your project (optional)

To avoid unnecessary Google Cloud Platform charges, use the Google Cloud console to delete your project if you do not need it.