How can admins review the performance of the Object Detection Model?

The performance of the Einstein Object Detection model is reviewed by analyzing its output, which is stored in the RetailVisitDetectedObject. Each time an image is processed, the model creates records in this object detailing the products it identified and their counts. Administrators can create reports, dashboards, and list views on the "Detected Objects" to compare the model's predictions against actual shelf conditions or manually entered data. This analysis provides direct, granular insight into the model's accuracy and effectiveness in real-world scenarios.

A. By creating a Custom Object: This is incorrect because Salesforce provides a standard object, RetailVisitDetectedObject, specifically for storing the model's results. Creating a custom object is unnecessary.

B. By enabling Custom Object Detection: This is a one-time setup step required to use the feature. It is not a method for monitoring or reviewing the model's ongoing performance.

D. Through Einstein Bots: This is incorrect as Einstein Bots are a separate conversational AI product used for customer service automation and are not related to image analysis or object detection.

1. Salesforce Help, Consumer Goods Cloud, Einstein Object Detection: "When a field rep uploads an image of a shelf, Einstein Object Detection identifies products, counts them, and checks planogram compliance. The results are saved to the RetailVisitDetectedObject." This confirms that the object stores the direct output needed for performance review.

2. Salesforce Help, Consumer Goods Cloud, Set Up Einstein Object Detection: This documentation outlines the setup process, which is distinct from performance monitoring. It details the permissions required to access objects like RetailVisitDetectedObject, implicitly confirming its role in the process.

3. Salesforce Developer Documentation, RetailVisitDetectedObject: The object reference for RetailVisitDetectedObject describes its fields, such as DetectedCount and ProductName, which are the core data points an administrator would use to build reports and analyze the model's performance.