Scientific Tropical Intelligence

See Beyond the Husk

AI-powered durian assessment for better harvest, grading, and market decisions.

Explore AIDurian →

Fresh durian fruits — the subject of AIDurian's AI assessment

1,216Puyat durian fruits in the primary field dataset

189Fruits in the structured multimodal collection

8NIR bands 720–980 nm

5AI assessment capabilities

2+1Platforms mobile · desktop · station

The Problem

The Challenge

Durian assessment is complex, subjective, and inconsistent — leading to postharvest losses, unfair pricing, and missed market opportunities.

Invisible Quality

Internal quality cannot be reliably seen from the outside.

Inconsistent Grading

Manual grading varies greatly between people and locations.

High Postharvest Loss

Poor decisions lead to losses in quality, value, and trust.

The Stakes

Why It Matters

Better assessment means better decisions — for growers, buyers, and consumers. AIDurian brings science and AI to every durian in the value chain.

See how AIDurian measures it →

How AIDurian Works

CAPTURE

Multimodal sensing captures RGB, multispectral NIR, thermal, and acoustic signals.

ANALYZE

AI models extract deep patterns across spectral, image, and sound features.

ASSESS

The system evaluates maturity, ripeness, defects, shape, and locules with confidence scores.

ACT

Insights support harvest timing, sorting, grading, and documentation decisions.

Five Assessment Capabilities

Five specialized AI modules — each validated on locally collected Puyat durian data.

Maturity Assessment

Multispectral NIR analysis of physiological maturity for harvest timing (immature, mature, overmature).

Ripeness Evaluation

Knock-sound analysis converted to spectrograms for eating-quality assessment.

Defect Detection

RGB image analysis that identifies and classifies surface defects and quality issues.

Shape Classification

Visual assessment of regular and irregular fruit morphology for grading support.

Locule Analysis

Segmentation and counting of locules from a top view of the fruit.

Platforms

Solutions

Purpose-built platforms for every part of the durian value chain — from field inspection to laboratory-grade analysis.

View all solutions →

INDAI Mobile

Field-oriented assessment in your pocket — image and audio capture with rapid quality feedback.

● Android App

DuDONG Desktop

GPU-supported analytics platform for parallel AI analysis, result history, and PDF reports.

● Windows Desktop

AIDurian Grading Station

Controlled imaging and acoustic environment for repeatable, standardized assessment.

● Research Prototype

Research & Results

Built on rigorous research with locally collected data. Every metric is reported exactly as validated — as documented in the AIDurian Terminal Report.

Explore research →

Maturity Classification

8-band NIR · 720–980 nm

84.37%mean accuracy · macro F1 83.47%

Acoustic Ripeness

Knock sound · spectrogram CNN

87.73%test accuracy · F1 89.29%

Defect Detection

RGB surface analysis

92.70%test accuracy · F1 88.30%

Locule Analysis

RGB top view

97.70%mAP50 · F1 98.69%

Shape classification additionally reached 91.10% mAP50 and 81.85% F1. Full dataset descriptions, test methods, and limitations are on the Research page.

Supported by