# Leaf Quality Grading - AI Grading from Any Smartphone

> AI-powered green leaf classification into Best / Below Best / Poor at intake. Works on any mobile device. No spectrometer required.

- Canonical URL: https://spectrify.ai/solutions/leaf-quality-grading

**Consistent Leaf Grading From Any Smartphone. No Device. No Delay. No Subjectivity.**

Classify green leaf intake as Best, Below Best, or Poor in seconds - consistent across every shift, every operator, every batch.

_TRI Validation in Progress_

- **Seconds**: Per grade decision
- **3 grades**: Best / Below Best / Poor
- **Any phone**: iOS & Android
- **Zero**: Hardware required

## The Problem

**The same lot gets three different grades depending on who's doing the check.**

Manual leaf grading is fast - but it's personal. Experienced supervisors disagree. Night shifts grade differently than day shifts. Suppliers learn who to present lots to. That inconsistency costs money.

- **Operator-dependent outcomes**: Manual grading varies 15–25% across operators in the same shift. A supplier who gets a poor grade on Tuesday can get a pass on Thursday with a different supervisor.
- **No audit trail**: When a buyer disputes a lot quality, what's your evidence? Manual grading leaves no record - just the memory of whoever was on shift.

## How It Works

**A photo from any mobile device is all it takes.**

1. **Photograph the intake batch**: The factory manager or intake operator photographs the green leaf batch using any iOS or Android device. No special setup required.
2. **AI classification**: The model classifies the lot into Best, Below Best, or Poor within seconds. Borderline cases are flagged with a confidence score for supervisor review.
3. **Record & store**: Grade result, photo, lot ID, timestamp, and operator are stored in the traceability dashboard - automatically, without any manual entry.

## Business Impact

**Consistency is worth more than speed.**

- **Shift-independent grading**: The same lot gets the same grade whether it's 6am or 11pm, Monday or Saturday. Operator skill and shift fatigue no longer determine quality acceptance.
- **Faster intake decisions**: Any operator can run the grade check - no need to wait for an experienced supervisor. Intake throughput increases, bottlenecks at peak flush reduce.
- **Dispute-proof records**: When a supplier disputes a lot rejection, the AI grade record - with photo, timestamp, and confidence score - is your evidence. No more word-against-word.

## Technical Specification

**System details**

| Specification | Detail |
| --- | --- |
| Input | Mobile camera photo of green leaf batch |
| Platform | iOS & Android - any modern smartphone |
| Grade outputs | Best / Below Best / Poor |
| Confidence scoring | Yes - borderline cases flagged for review |
| Validation body | TRI validation in progress |
| Optimised for | Sri Lankan green (fresh) leaf at factory intake |
| Records stored | Photo, grade, confidence, lot ID, timestamp, operator |
| Dashboard | Traceability dashboard - included with all plans |

## Also in the Platform

- [Moisture Detection](https://spectrify.ai/solutions/moisture-detection): TRI-validated NIR moisture in 15 seconds. The only validated meter in Sri Lanka. (TRI Validated, December 2025)
- [TPP Detection](https://spectrify.ai/solutions/tpp-detection): Total polyphenol measurement in 15 seconds. R² > 0.93 accuracy. (LKR 15, vs LKR 6,000 lab test)
- [Traceability](https://spectrify.ai/solutions/traceability): Every reading linked to lot, time, and operator automatically. (Included, with all plans)

## See It in Action

[Watch a live demo of SpectrifyAI in the field.](https://www.youtube.com/watch?v=gqS4Vd871A4)

## Common Questions

### Does it work on processed tea or only green leaf?

The model is trained and optimised for green (fresh) leaf at factory intake - which is where grading decisions are made and disputes originate. Processed/made tea grading is a different model; contact us if that's your use case.

### How consistent is AI grading compared to an experienced supervisor?

Manual grading typically varies 15–25% across operators in the same shift, and more significantly across different supervisors and time of day. The AI model applies the same criteria on every photo. Consistency - not just accuracy - is the primary value.

### What if lighting conditions at our intake area vary?

The model is trained with lighting variation in mind. Standard factory intake lighting is sufficient. Extremely low-light environments or direct outdoor sunlight can affect results - we recommend a consistent, diffuse lighting setup, which most factory intake areas already have.

### What happens to grades that fall on the boundary between Best and Below Best?

The model outputs a confidence score alongside the grade. Borderline cases - where the model is less certain - are flagged for supervisor review. This is by design: borderline lots get human attention, not just a binary label.

### Does the grading result link to the traceability dashboard?

Yes. Every leaf grade result is automatically stored with lot ID, timestamp, and operator - the same as moisture and TPP readings. The result is accessible in the dashboard and can be exported with the full lot report.

## Start Grading Consistently - No Hardware Needed.

Works with any smartphone camera at intake. Book a demo and see a live grading run on your sample batches.

[Book a Demo](https://spectrify.ai/contact)