A Consumer Goods Cloud user can review vision detection accuracy through two primary mechanisms. Model Metrics provide a technical assessment of the underlying Einstein Vision model's performance, using standard machine learning metrics like F1 Score, Precision, and Recall. This allows administrators to evaluate the overall health and accuracy of the AI model itself. Shelf Metrics, which are the Key Performance Indicators (KPIs) generated after an image is analyzed (e.g., Share of Shelf, Facing Count, Out-of-Stock), allow field reps and managers to review and validate the accuracy of the detection for a specific shelf image taken during a store visit.

A. Goods Metrics: This is not a standard term used within Salesforce Consumer Goods Cloud for evaluating the accuracy of the Einstein Vision model or its outputs.

C. Planogram Metrics: This typically refers to the planogram compliance score, which is an outcome of the vision analysis, not a direct measure of the object detection model's accuracy.

D. Object Metrics: This is a generic term. The specific and correct terminology used in Salesforce for evaluating the AI model's performance is "Model Metrics."

1. Salesforce Help Documentation, "Monitor Your Einstein Vision Model": This document details the process of evaluating model performance. It states, "Use the Model Metrics tab on the Einstein Vision Model record page to see how your model performs. The tab shows metrics such as F1 Score, Precision, and Recall for the model." This directly supports the correctness of 'Model Metrics' (E).

2. Salesforce Help Documentation, "How Image Recognition Works": This document explains the output of the image analysis process. It states, "The service analyzes the image and identifies the products on a shelf... The results are saved as retail store KPIs. For example, share of shelf, out-of-stock, or planogram compliance." These KPIs are the Shelf Metrics (B) that a user reviews to confirm the accuracy of a specific scan.