Use OCI Vision Models in Oracle Analytics

Use prebuilt OCI Vision models to build image recognition and text recognition into your applications without machine learning (ML) or artificial intelligence (AI) expertise.

For example, you might use object detection to identify cars in photographs, or detect faces in photographs so that you can blur them to protect peoples' identities.

Pretrained OCI Vision models available include:

• Pretrained Image Classification.
• Pretrained Image Face Detection.
• Pretained Image Text Detection.
• Pretrained Object Detection.

If you have more than 20,000 images to process, in OCI's Object Storage & Archive Storage area you typically set up multiple buckets containing no more than 20,000 images in each bucket. Then, you create a separate data flow to process each bucket, and use a Sequence to sequentially process multiple data flows so that you can process all of your images.

Prerequisites:

• Make sure that your Oracle Analytics instance is integrated with OCI Vision. See Integrate Oracle Analytics with OCI Vision.
• Prepare a dataset that references the images that you'd like to analyze and upload it to Oracle Analytics. See Prepare Images for Analysis With an Vision Model.

1. On the Oracle Analytics Home page, click Create, and then click Data Flow.
2. Select the dataset linking to the images you want to analyze, then click Add.
3. In the Data Flow editor, click Add a step (+).
4. From the Data Flow Steps pane, double-click Apply AI Model, and then select the model to use.
   For example, you might select "Pretrained Object Detection" to detect car number plates. Or, to detect faces in photographs, you might select "Pretrained Image Face Detection".
5. In Apply Model, go to the Parameters section, and configure the Input Column and Input Type parameters.
   • If you're referencing your source images by bucket, in Input Column select URL, and in Input Type select Buckets.
     
     
     Description of the illustration vision_parameters.png
     
     
   • If you're referencing your source images individually, in Input Column select File Location, and in Input Type select Images.
   See Parameter Options for Object Detection, Image Classification, and Text Detection Analysis Models.
6. Optional: Use the Inputs and Outputs options to change the default settings (the options available depend on the model type).
   
   
   Description of the illustration oci-vision-face-recognition.png
   
   
7. In the data flow editor, click Add a step (+) and select Save Data.
8. Enter the name of the dataset in which to store the output results.
   For example, you might call the dataset 'Car Parking Analysis Results'.
9. In the Save data to field, specify the location for saving the output data.
10. Click Save, enter a name and description for the data flow, and click OK to save the data flow.
11. Click Run Data Flow to analyze the images and output the results in a new dataset.
    If you have fewer than 20,000 images, you can process them in a single data flow. If you have more than 20,000 images, create a separate data flow to process each bucket (that is, using a separate dataset for each bucket), and use a Sequence to sequentially process multiple data flows. After you've created multiple data flows, on the Oracle Analytics Home page, click Create, and then click Sequence.

When the data flow completes the analysis, open the dataset that you specified in Step 7.

To locate the generated dataset, from the Oracle Analytics home page, navigate to Data, then Datasets.

Description of the illustration vision_output_ds.png

For more detail about the generated results, see Output Data Generated for Object Detection, Image Classification, and Text Detection Analysis Models.