Let's build a backend workflow that downloads videos from an AWS S3 bucket, detects objects on these videos (TensorFlow algorithm) and uploads the tagged video back to the S3 bucket.
So let's build this workflow with Ryax.
We login and create a new workflow.
We can choose from a wide variety of triggers to start the workflow.
We select a single execution trigger for now.
You can configure each workflow module in the right panel.
Now we add an AWS S3 bucket dlownload action.
And we set it up using our bucket credentials.
Now we can work on the video itself.
We cut the video frame by frame so the algorithm can detect objects on single images.
We use the output from the previous action (the video) and ingest it into the next action via a 'link' path.
Next we feed the individual frames to the TensorFlow object detection algorithm.
We choosed th MobileNet v2 model.
And link the previous output to his action.
Last step is to reconstruct the result video with the tagged frames.
We're now ready to deploy our workflow.
We can also generate an API on the spot to expose any of our workflow steps as API endpoints.
Let's deploy our workflow now.
Once deployed our workflow enters a ready state where it can be triggered and gereate executions.
Let's take a look at our first run.
This view allows us to observe the workflow while it's running.
We are able to access timestamps, logs and intermediate data for all our workflow's steps.
We download all intermediate data to check it out.
We can check these out for errors and status reports.
Now let's go back to our workflow and add a bucket upload step.
We configure it in a similar fashion as the existing download step.
Now let's deploy our workflow again.
And observe the steps running.
This time our result video has been uploaded back to the bucket.
This backend workflow is now ready for production.