Version 3.2.5
Subminor Version - Released on Nov 16, 2025
1. Peaberry-Specific Density Tool
As Peaberry beans have a relatively higher density compared to regular beans, a dedicated input field was added so users can define a custom density value for peaberries, overriding the standard class-density maps.
Navigate to "Edit Model" in the side menu and select the desired AI model.
Scroll down to the "Weight Estimation" panel.
Click on the "Include Peaberry Density" switch
(1).Add the density value for peaberries by typing it directly
(2), or import an .hdf5 file with a known sample weight to calculate the density automatically.Click on "Save" to store the new data.
During the weight estimation calculation, every image with an index greater than or equal to the start of the peaberry record will use the peaberry density instead of the class-specific density.
Changing the class of a peaberry-indexed image will not affect the final sample weight, since all peaberry seeds, regardless of classification, use their screen-specific density for the weight calculation instead of the class-based density
2. Group Creation in Point-Based Methods
Groups can be created for any grading method.
For example, in this COB methodology for Brazil Arabica coffee, four groups were defined to represent relevant information that QC labs typically check manually.
Each group was created by selecting classes according to the user’s criteria and applying a screen-size rule of screen equal or greater than 12.
A group named PVA was created using the classes at indexes 0, 12, and 14, and applying the rule to include only seeds with screen size greater or equal to 12
(1).A group for smaller seeds (less than or equal to 12) was also created by selecting all classes at once.
Groups can include repeated classes, and there are no restrictive rules for creating them.
2.1 Visualizing Groups in the Dashboard
Groups can be visualized in the Dashboard under the "Classification Methods" panel.
3. Multi-Analysis Images Export
"Export Images" menu now allows for multi-analysis export.
To change the mode, click on the "Single/Batch" switch on the top left of the panel.
"Single" will export only the opened file.
"Batch" will export all .hdf5 files present in the selected folder.
"Group Analysis" check box will group the export from all analyses into a single folder.
"Apply Similarity" will run a clustering algorithm to find similar images and group them together.
Using "Group Analyses" and "Apply Similarity" on a batch of analyses will create folders with similar images from multiple analyses, acting as a pre-classifier for images and automating the process of creating and cleaning data for dataset preparation.
4. Multi-Analysis Classification Report export
Similarly, the "Classification Report" also supports multi-analysis export.
To change the mode, click on the "Single/Batch" switch at the top left of the panel.
"Single" will export only the file currently open.
"Batch" will export all .hdf5 files located in the selected folder.
5. Multi-Analysis Excel and CSV export
Similarly, the "Export Data" also supports multi-analysis export.
To change the mode, click on the "Single/Batch" switch at the top left of the panel.
"Single" will export only the file currently open.
"Batch" will export all .hdf5 files located in the selected folder.
6. Screen-Size-Per-Class Plot
A new plot showing the class distribution across screen sizes has been added to the dashboard.
This visualization highlights how each defect type is distributed along the screen-size spectrum, providing analytical insight into defect occurrence patterns and supporting more informed operational or quality-control decisions.
Last updated