Currently, we're set up only for sending image annotation requests for a single image at a time.
Google Cloud Vision also provides for asynchronous batched requests of up to 2000 images at a time, as described at https://cloud.google.com/vision/docs/batch. We should update GoogleCloudVisionHandler to be able to use it. I anticipate that all of our annotation requests will ultimately use the batch API.
Note: This will require some additional setup, namely to set up a Google Cloud Storage bucket to receive results, and to configure our Handler to retreive results from it.
Open questions
- What happens in case of error retrieving or labeling an image during a batched request?