Domain

Measurement & Evidence

Methods and systems that generate reliable, scalable measurements from field imagery, sensors, and satellite data — turning physical-world observations into evidence for policy and programme design.

6applications

33technologies

Applications

BarkVisionAI

CNN bark-image species ID and large-scale benchmark dataset, published in Nature Scientific Data.

156,001Images

Nature Sci. DataVenue

CNNResNet50Transfer learningImage classification

HaatVisionAI

Smartphone CV pipeline for reference-free 3D food volume estimation at rural market scale.

95%Classification

14%Volume error

Mask R-CNNResNet18LSTM6DoF pose

MonoDBH

Monocular DBH estimation from a single RGB image via GroundingDINO, SAM 2, and pinhole projection.

0.953Mask mAP50

1 RGB frameInput

GroundingDINOSAM 2Pinhole projectionObject detection

AGB Estimation

draft

Above-ground biomass from DBH, species, and height via species-specific allometric equations.

DBH · spp · HInputs

AllometricMethod

Allometric modellingMonoDBHBarkVisionAI

Forest Resource Inventory Mapping

Multi-sensor species distribution modelling across landscape-scale forest regions.

S1 + S2Sensors

Random ForestModel

SDMRandom ForestSentinelGeospatial rasters

Income Baseline

draft

Multi-source farmer income estimation fusing mobile, satellite, and field imagery.

Data fusionApproach

In progressStage

LVLMNighttime lightsVision modelsStratified sampling

Image Data Cleaning Pipeline

Two-phase dedup and multi-LLM label validation for crowdsourced survey imagery.

15,197Images

29.6%Dupes removed

pHashHSVMulti-LLM consensusLVLM

Measurement & Evidence

Focused · click to expand