Data collection

We insist that, before starting to collect data, all products should be arranged with the main part facing forward. If it is not in this format, you may have a bad recognition analysis by the system. It is not recommended to capture blurred images or in bad lighting conditions.

It is recommended to capture the area of interest in a single photo, thus ensuring good resolution.

When it is not possible, in a single capture, to record the precise product, we recommend using the crop tool to select the area of interest in the image.

We recommend not capturing photos of the products from the side, preferring to capture the products directly from the front. When not feasible, use the crop tool to outline the area of interest and correct the perspective.

It is not recommended to capture just one or a few products. Automatic recognition works best when there are more contexts in the image.

Data editing

Edit misidentifications in reports by adding, removing or changing products. It is important to note that the result depends on the planogram used, and if the product is not in the planogram, it will not be added to the report.

The edits or overlays must cover the products properly, that is, completely.

Currently, we have two types of annotations. The first annotation is performed by the field operators and the second is performed by the data supervisors, after the synchronization of the field operators.

It is important to emphasize:

- That the data supervisor should be informed when there is an incorrect planogram, such as missing the flavor of a specific product.

- That the data supervisor must be informed about the lack of an important product when trying to edit the report, as, the specific flavor is not registered.

- That the data supervisor should be informed about identifying duplicate or confusing names or thumbnails.