Hello video data: Deploy a model to make batch predictions

This tutorial has several pages:

  1. Setting up your project.

  2. Creating a video classification dataset.

  3. Training an AutoML video classification model.

  4. Deploying the model for batch predictions.

  5. Cleaning up your project.

Each page assumes that you have already performed the instructions from the previous pages of the tutorial.

The email you receive informing you that Vertex AI has completed training your model has a link to your models: Vertex AI dashboard

Select the model to continue.

1. Create batch predictions

  1. Select the Batch predictions tab in the model page, then select Create batch prediction.

    Vertex AI dashboard

  2. The New batch prediction form appears. You will be providing a list of videos (source file) to be classified.

    • Provide a Batch prediction name.
    • For the Source path use [gs://]automl-video-demo-data/hmdb_split1_predict.jsonl
    • Provide the Destination path to your bucket. Results will be added to the predict_results folder
      [gs://]your-own-gsc-path/predict_results
    • Select Create.

    Vertex AI dashboard

View results

Once the job is done, demo_data_predictions will show up under the Batch prediction tab.

  1. Click the demo-model-prediction link in the Batch predictions view.

    Vertex AI dashboard
    Details of the batch prediction job appears.

  2. Click the link to the right of Export location to view the results in your storage bucket.

    Vertex AI dashboard
    The results in your storage bucket appear.

    Vertex AI dashboard

  3. To see your results in the UI, click View results.

    Vertex AI dashboard
    A video appears.

    Vertex AI dashboard
    From the drop down, you can select other videos you want to see the results for.

    Vertex AI dashboard
    You can also adjust the confidence threshold to hide/show the predictions.

    Vertex AI dashboard

View the results

In the results for your video annotation, Vertex AI provides three types of information:

  • Labels for the video. You can find this information under the Segment tab below the video on the results page.
  • Labels for shots within the video. You can find this information under the Shot tab below the video on the results page.
  • Labels for each 1-second interval within the video. You can find this information under the 1 Second Interval tab below the video on the results page.

If you want to see more labels, you can change the threshold score when you request a prediction. AutoML video only displays the labels that are above the specified threshold.

If the prediction fails, the results in the list show a red icon on the Recent Predictions list.

If only one video in the prediction attempt failed, the prediction shows green in the Recent Predictions list. On the results page for that prediction, you can view the results for the videos that Vertex AI has annotated.

What's next

Follow the last page of the tutorial to clean up resources that you have created.