Version 3.2.1

Subminor Version - Released on Sep 21, 2025

1. Sieve Offset

Sieve Offset is a feature designed to override the Image Scale, providing a separation between density calculations and screen-size distributions.

• Click “Camera On” (1) to power on the camera.

• Click “Light” (2) to turn on the illumination.

• Position the checkerboard, as shown in the image below, so that all squares are fully visible in the camera view.

• Click “Use Checker Board” (4) to begin the calibration process.

• A color grid will appear, and the system will automatically perform the measurements.

• After finishing, the new Image Scale value will be filled automatically.

• If any error occurs, make sure all checkerboard squares are visible and try again.

• Click “Save Settings” (5) and then “Camera Off” to complete the process.

• "Image Scale" is now automatically obtained from the Checkerboard, and it is used to calculate the density based on the real image scale.

• Check the methodology for weight estimation based on density approximations in the link.

• "Sieve Offset" is the value that will override the "Image Scale" to fit the screen size distribution.

• As the "Screen Size" is not a physical measurement but rather a mechanical optimization, adjustments need to be performed to match real-world equipment.

Image Scale (pixel/cm) is the variable used for all calculations related to weight estimation. It defines how the 2D area, measured in pixels², is converted into cm².

Sieve Offset (pixel) is the variable applied to fine-tune the effective Image Scale in order to match the physical screen-size distribution of real equipment.

2. Weight Management Panel

The Weight Management panel is designed for the user to visualize and, when necessary, adjust the estimated weight of the sample.

• To access the Weight Management panel, click on “Inspected Seeds.”

• The sample weight is displayed in the input field (1).

• If the estimation is significantly incorrect, the user can redefine the weight by typing a new value and clicking “Update Weight”.

• This action saves the new sample weight and rescales all class densities relative to one another, ensuring the final estimated weight matches the desired value while preserving the proportional differences between classes

• In the example above, the user changed the weight from 106.1 g to 105 g.

• This adjustment affects the weight factor (the multiplier used to reach the required method weight) and all subsequent calculations, including equivalent defect recalculations.

• A flag labeled “User Input” (2) is displayed in the panel to indicate that the weight was manually entered rather than estimated.

• To return to the original estimated weight, simply click “Reset Weight”.

• At the bottom of the panel, it is possible to enable or disable the log error linked to the sample weight.

• As stated by Csmart`s SOP, the minimum expected weight is 300g.

• Samples with lower weights produce higher errors due to reduced statistical representativeness.

• The log error range tied to the weight range is shown in the table, and the error can be displayed together with the equivalent defects if the user requires it.

The Weight Management Panel allows users to review and adjust the sample weight used in density-based calculations. If needed, the weight can be manually corrected, triggering automatic rescaling of all class densities and recalculation of equivalent defects. The system also shows whether the weight was estimated or manually entered, and provides an overview of expected log error ranges based on the sample’s total weight.

3. Company Logo in Classification Report

Add your company logo to the PDF reports by selecting an image and enabling the logo option.

• It is now possible to include a company logo in the Classification Reports PDF.

• Click on "Select Image" (1) and choose the logo to be used.

• Enable the "Include company logo" switch (2) to add the logo to the final export.

• Once selected, the image path will be saved and reused automatically in future exports.

4. Peaberry Record Mode

Peaberry can be analysed in the Dashboard with the new tool to record its specific screen sizes. Peaberries are recorded separately from the regular seeds within the same analysis file. After finishing the regular portion, keep the session open, switch the Seed Type to “Peaberry,” and pour only the peaberry fraction so the system records the MK-sieve distribution correctly. This ensures both portions are processed independently and combined accurately in the final analysis.

• Csmart incorporates a module to record and present peaberry beans and their respective screens.

• Due to limitations of the technology, peaberries cannot be detected by AI using 2D images. As an alternative, the following approach was defined:

  • Manually divide the sample into two groups: Regular Beans and Peaberry Beans.

  • This division is done by screening the sample using lab sieves in the following order:

    Low Grade | MK09 | Screen 15 | MK 10 | Screen 16 | MK 11 | Screen 17

  • Peaberry (Moka) sieves are grouped together, while all regular sieves (including the low-grade ones) form the second group.

  • The software allows users to record these two groups at separate moments, as described below.

• Navigate to “General Settings” in the side menu.

• Scroll down to Dashboard and Report Settings and enable the “Peaberry Record Mode” and “Display Peaberry Sieves” switch buttons (1).

• These changes need to be done only once, as they will persist in future software usage.

4.1 Recording a New Analysis with Peaberry Beans

• When recording a new analysis, a new tool indicating "Seed Type" is presented in the record panel.

• The recording always starts with the regular portion of the sample (flat seeds).

• Once all regular (flat) seeds are recorded, without stopping the session, simply click the "Regular" switch to continue.

• Now the Seed Type will indicate "Peaberry", and the peaberry portion of the sample can be poured into the hopper.

• The recording continues normally until the entire sample has been processed.

• The file now stores an index indicating the seed number at which the peaberry portion begins, dividing the sample into two groups.

• The AI classification will run normally for the peaberry portion, since this process is based on screen-size classification rather than defect categories.

• Outcomes such as immature-peaberry, ok-peaberry, or sour-peaberry are fully supported and will occur under this new approach.

Enable the Peaberry Record Mode and Visualization once in General Settings, and the software will automatically process both portions of the sample in the final report.

Peaberry recording is done in two parts: record the regular seeds first, then switch to Peaberry mode without stopping the analysis.

4.2 Visualizing an Analysis with Peaberry Screens

• The dashboard now presents the Peaberry screens in new check boxes under the regular screens.

• MK09, MK10, and MK11 can be selected all at once or in any order.

• The average screen size calculation is also affected by the inclusion of the Peaberry sieves, using their numeric values as factors: MK9 is counted as 9, MK10 as 10, and MK11 as 11.

The Peaberry (Moka) screens work as real sieves. If no peaberry screen is selected, the seeds remain in the appropriate regular screens. Including or removing the Peaberry screens does not remove the seeds, only changes how their screen sizes are classified.