An Einstein object-detection implementation depends on three essential elements: a suitable training dataset, a properly trained and evaluated model, and appropriate user access to the functionality.
The implementation begins by collecting representative images of the products or objects that must be detected. For object detection, the training images require labeled bounding boxes identifying the position and dimensions of each object. These examples are uploaded as a dataset and used by the training process to create an object-detection model.
The trained model should then be evaluated before deployment. Salesforce ' s model-training architecture reserves data for testing and provides model information that allows implementers to evaluate the model ' s detection performance. Reviewing model quality before exposing the capability to field representatives is essential because representative training data and adequate model performance directly affect detection results.
Finally, the field users who execute the functionality require the relevant access and permissions. Permissions are part of the Salesforce security model and determine whether users can access the required Consumer Goods and Einstein functionality.
OCR configuration and competitor-suppression datasets are not baseline requirements for this object-detection setup.
Study Guide Topics:Platform - > Einstein Vision - > Object Detection - > Dataset - > Model Training - > User Permissions
===============