01 · Sample overview
UG-2026-0142
Sample weight150 g
Analyzed beans1,674
Green coffee grading,
bean by bean.
360° view of every bean Up to 500 g sample Arabica and Robusta SCA, COB, NY, GCA
/complete workflow
Watch the sample become a record.
Register the sample, load the beans, follow the scan, and open the finished result.
-
01Register
Sample ID, origin, processing method and weight go in on the touchscreen or in the web application.
-
02Load
The beans go in. VeroLab separates them and captures a 360° view of every one.
-
03Review
The finished record opens on any screen: sizes, defects, colour, and the beans behind each number.
/from sample to bean Start with the sample.
Start with the sample.
Inspect down to the bean.
This is a complete VeroLab record: sample context, screen sizes, every classified defect, the beans behind each number, and the colour analysis.
02 · Screen sizes
Size distribution by count and weight.
See where the sample sits across standard screens and separate good coffee from classified defects.
14.69mean screen
<13
13
14
15
16
17
18
19
20+
| Screen | Beans | Mass | All | Good | Defects |
|---|---|---|---|---|---|
| 20+ | 26 | 4.19 g | 2.80% | 1.05% | 1.75% |
| 19 | 51 | 7.24 g | 4.83% | 2.96% | 1.86% |
| 18 | 101 | 12.92 g | 8.62% | 5.56% | 3.05% |
| 17 | 146 | 17.06 g | 11.37% | 6.71% | 4.66% |
| 16 | 225 | 23.52 g | 15.68% | 9.78% | 5.90% |
| 15 | 216 | 20.35 g | 13.57% | 7.42% | 6.15% |
| 14 | 251 | 21.09 g | 14.06% | 7.44% | 6.62% |
| 13 | 214 | 16.36 g | 10.91% | 4.76% | 6.15% |
| <13 | 444 | 27.26 g | 18.17% | 4.89% | 13.28% |
| Total | 1,674 | 150.00 g | 100% | 50.57% | 49.43% |
03 · Defects SCA
Every visible defect, quantified.
Counts are translated into sample share, estimated weight, and Full Bean Equivalent for review against SCA guidance.
52.09%
515total FBE
| Defect | Count | Share | Weight | FBE |
|---|---|---|---|---|
| Full black | 386 | 23.06% | 30.29 g | 386 |
| Full sour | 4 | 0.24% | 0.40 g | 4 |
| Foreign matter | 10 | 0.60% | 1.21 g | 10 |
| Cherry pod | 13 | 0.78% | 1.52 g | 13 |
| Part black | 98 | 5.85% | 9.36 g | 32 |
| Withered | 6 | 0.36% | 0.58 g | 1 |
| Shell / ear | 19 | 1.14% | 2.00 g | 3 |
| Broken / chipped / cut / flat | 229 | 13.68% | 18.85 g | 45 |
| Parchment | 15 | 0.90% | 2.11 g | 3 |
| Hull / husk | 92 | 5.50% | 7.82 g | 18 |
| Total defects | 872 | 52.09% | 74.14 g | 515 |
04 · SCA defect images
Move from a number back to the beans.
The report’s dalvany image panels display the original classified bean tiles with their confidence scores.
Original classified beans from UG-2026-0142, ordered by confidence. Hover to inspect. Click a bean to open its captured views.
Good802 classified beans · 8 shown
dalvany-image-panelFull Black386 classified beans · 8 shown
dalvany-image-panelFull Sour4 classified beans · 4 shown
dalvany-image-panelCherry / Pod13 classified beans · 8 shown
dalvany-image-panelForeign Matter10 classified beans · 8 shown
dalvany-image-panelPart Black98 classified beans · 8 shown
dalvany-image-panelInsect Severe0 classified beans · 0 shown
dalvany-image-panelNo images.
Mold0 classified beans · 0 shown
dalvany-image-panelNo images.
Immature0 classified beans · 0 shown
dalvany-image-panelNo images.
Withered6 classified beans · 6 shown
dalvany-image-panelFloater0 classified beans · 0 shown
dalvany-image-panelNo images.
Shell / Ear19 classified beans · 8 shown
dalvany-image-panelBroken / Chipped / Cut / Flat229 classified beans · 8 shown
dalvany-image-panelParchment15 classified beans · 8 shown
dalvany-image-panelLight Insect0 classified beans · 0 shown
dalvany-image-panelNo images.
Hull / Husk92 classified beans · 8 shown
dalvany-image-panel 05 · Colors
Average colour with its underlying distribution.
The live report combines its RGB histogram and CIELAB plot with the calculated average colour.
Average RGB#4F4734
R79
G71
B52
/360° analysis
360° can change the answer.
Beans move through the imaging area one at a time, so no two overlap and the conditions stay the same from sample to sample.
A bean may appear good from one side and clearly defective from another. The AI model evaluates each view separately before the predictions are combined.
This is particularly important for defects such as insect damage, shells, chips, broken beans, and part-black beans.
- Example 01Final label
Part Black
The black area is only visible from one side. The strongest view carries the final label.
Closest alternative: Full Black
Real VeroLab bean captures. View scores shown for illustration. - Example 02Final label
Shell / Ear
The hollow shell shape is clear from the front view and confirms the label over a broken-bean reading.
Closest alternative: Broken
Real VeroLab bean captures. View scores shown for illustration. - Example 03Final label
Broken / Chipped / Cut / Flat
A chipped edge appears in two of three views, so the aggregation keeps the defect label.
Closest alternative: Shell / Ear
Real VeroLab bean captures. View scores shown for illustration. - Example 04Final label
Withered
Scores stay close across views. The bean is labelled Withered but flagged for human review.
Closest alternative: Immature
Real VeroLab bean captures. View scores shown for illustration. - Example 05Final label
Full Sour
Color alone is ambiguous. The final label is applied with visible uncertainty instead of a hidden decision.
Closest alternative: Full Black
Real VeroLab bean captures. View scores shown for illustration. - Example 06Final label
Cherry Pod
All three views agree. The dried cherry is counted under the protocol’s Cherry Pod category.
Closest alternative: Full Black
Real VeroLab bean captures. View scores shown for illustration. - Example 07Final label
Parchment
The parchment layer is most visible from the third view, which separates it from loose hull fragments.
Closest alternative: Hull / Husk
Real VeroLab bean captures. View scores shown for illustration.
/how the model decides 360° views merge into one result.
01 02 03
360° views merge into one result.
You set the rules.
View-level prediction
The AI model evaluates each image and produces confidence scores for possible defect categories.
Aggregation and Thresholds
The aggregation layer combines the predictions and applies the agreed thresholds and rules.
Final label
The bean receives its final defect label, while the underlying views and confidence scores remain available for review.
A confidence score describes how strongly a view resembles a learned category. It is not an accuracy percentage for the complete sample.
/grading logic /layer 001 /layer 002 /layer 003
One AI model. Made to fit your operation.
Thresholds, precedence and severity rules are set with your quality team during onboarding, so the same model reports the way your operation grades.
AI model
Recognizes visible characteristics and possible defect classes.
Grading protocol
Determines how defects are named, counted, and converted into Full Bean Equivalent.
Current protocols
SCACOBNew YorkGCA
Grading profile
Applies agreed thresholds, precedence rules, severity rules, or internal specifications for a particular origin, process, product, or methodology.
/coffee variation
Built for variation, not one perfect sample.
Supported as standard
ArabicaRobustaSpecialty gradesCommercial gradesWashed coffeeNatural coffee
May require additional alignment
HoneyAnaerobicStrong fermentationUnusual color profilesDecaffeination workflowsInternal standardsAdditional origin-specific defects
These coffees are not rejected as a category. Representative samples may be needed to determine whether the requirement can be handled through grading alignment or requires custom work.
/sample comparison
Follow the coffee from offer to arrival.
Compare linked samples at different points in the supply chain.
Typical comparisons include:
- Offer, pre-shipment and arrival samples.
- Current and previous crop.
- Before and after a defined processing step.
The comparison can show changes in screen sizes, visible defects, estimated weights, and color.
/web application and API Use the web application.
/software 001 /software 002 /software 003
Use the web application.
Or take the data with you.
Standalone use
Register samples, review results, inspect beans, and generate reports in the VeroLab web application.
Exports
Export individual results as PDF and multiple-sample data as CSV.
Standard API
Retrieve VeroLab results into your ERP, CRM, Cropster environment, or internal system.
/beyond green grading /module 001 /module 002 /module 003 /module 004
Green grading is the starting point.
The core product does not require additional software modules. Optional modules use green coffee data in other parts of the operation.
Cupping
Record sensory results in the same sample record. Human cupping remains necessary.
Status: In developmentBlending
Visualize combinations of available lots against screen-size, defect, sensory, and financial targets.
Status: In developmentTrading
Quality-to-price analysis and benchmarking.
Status: In developmentRoasting
Use green coffee grading information in roasting-related workflows.
Status: In development /technical details 
Built to fit a Q Lab.
| Specification | Current product |
|---|---|
| Dimensions | 38 × 34 × 24 cm |
| Weight | 15 kg |
| Sample range | 1 g to 500 g |
| Recommended sample | 300–350 g |
| Typical analysis time | ~5 minutes for a standard sample |
| Lower-grade samples | ~7–8 minutes |
| Imaging | 360° views per bean |
| Coffee | Arabica and Robusta |
| Grades | Specialty and commercial |
| Standard processes | Washed and natural |
| Protocols | SCA, COB, NY, GCA |
| Color | RGB and CIELAB |
| Volume | Calculated |
| Bean weight | Estimated |
| Density | Estimated |
| Moisture | Not measured |
| Water activity | Not measured |
| Processing | Locally on the device |
| Connectivity | Wi-Fi |
| Voltage | 110 V and 220 V |
| Included accessories | Scale and barcode scanner |
| Installation | Place on a workspace, connect power and Wi-Fi; setup is supported remotely |
| Additional equipment | None needed |
| Cleaning | Automated cycle after every sample |
| Maintenance | Your team cleans the relevant compartments, roughly every two to three months depending on sample volume and dust |
| Support | Remote troubleshooting first; if that cannot resolve it, the device is serviced in Berlin under the applicable service and warranty terms |
| Physical sorting | Not included |
/your samples
See what VeroLab does with your coffee.
Request an offer and send up to five representative samples for validation.