1Z0-184-25 Question 149
Select 2Your company wants to develop a real-time image classification solution using Oracle Cloud Infrastructure (OCI) Vision. They plan to store user-uploaded images in OCI Object Storage and require a custom-trained model for specialized image categories. They also want to ensure that the model can be easily invoked via an API endpoint for real-time predictions. Which two steps must you take to successfully implement this pipeline?
- A
Create a dataset in OCI Data Labeling, label the uploaded images, and train a custom model with OCI Vision based on that labeled dataset.
- B
Deploy the custom model using OCI Functions to create a serverless real-time classification endpoint.
- C
Deploy the trained model as a hosted endpoint within OCI Vision, enabling real-time inference through a published API.
- D
Configure an OCI Event Rule to automatically re-train the model whenever new images are uploaded to OCI Object Storage.
- E
Use Oracle Data Safe to store and manage the labeled images before training the model in OCI Vision.
Show answer and explanation
Correct answers: A, C
Explanation
To implement a real-time image classification pipeline in Oracle Cloud Infrastructure, you must label your images using OCI Data Labeling and then train a custom model with OCI Vision. Once trained, the model should be deployed as a managed endpoint in OCI Vision, enabling REST API access for real-time predictions. Refer to Oracle� documentation on OCI Vision (https://docs.oracle.com/en-us/iaas/Content/vision/home.htm) and OCI Data Labeling (https://docs.oracle.com/en-us/iaas/Content/dat-labeling/home.htm) for best practices on setting up and optimizing your AI workflow.
- A. Correct.
Correct. You must first label your images using OCI Data Labeling so that OCI Vision can learn from labeled examples. OCI Vision supports creating custom models trained on these labeled datasets, allowing for specialized classification tasks.
- B. Incorrect.
Incorrect. Although OCI Functions can host custom code, OCI Vision models are not generally deployed through Functions for real-time inference. OCI Vision itself provides a native service for model deployment, allowing direct API integration.
- C. Correct.
Correct. OCI Vision enables you to deploy your trained model as a managed endpoint. This option gives you a low-code way to expose the model for real-time predictions without setting up additional infrastructure.
- D. Incorrect.
Incorrect. While you could automate certain workflows with OCI Event Rules, automatically re-training the model on every new image is not typically required or recommended. Model re-training is usually done periodically or when sufficient new labeled data is available.
- E. Incorrect.
Incorrect. Oracle Data Safe is designed for securing and auditing databases, not for storing or labeling images. OCI Object Storage and the Data Labeling service are the correct components for handling image data for training.