CSIRO: Image2Biomass: AI for Precision Biomass Prediction

Opportunity for

• AgTech companies, industry partners, and investors interested in commercialising AI and computer vision technologies that enable accurate, species level pasture biomass estimation

Opportunity description

Industry challenge

Grazing systems cover approximately 50 percent of Australia’s landmass and around 25 percent of the Earth’s land surface globally. Accurate measurement of pasture biomass is critical for effective grazing management, productivity, and long term environmental sustainability. However, accurately estimating pasture biomass at scale remains a persistent challenge.

Pasture management is a complex optimisation problem. Producers must balance livestock demand with pasture growth potential, species composition, climate variability, and management practices to achieve both short term productivity and long term land health. Biomass quantity alone is insufficient. Species level information is essential to assess pasture quality, regrowth potential, and nutritional value.

Existing measurement methods are limited. Clip and weigh approaches are accurate but slow and impractical at scale. Plate meters and capacitance meters are faster but unreliable under variable pasture conditions. Remote sensing enables broad coverage but cannot reliably distinguish biomass at the species level and still requires manual calibration.

This creates a clear opportunity for scalable, accurate, and low friction biomass measurement that can operate at paddock scale while delivering species level insight. Image-based AI models provide a pathway to overcome current trade offs between accuracy, cost, and usability.

Current opportunity

CSIRO, Australia's national science agency, in collaboration with Meat and Livestock Australia and Google, launched a global Kaggle competition in October 2025 to advance species level pasture biomass prediction using images. The competition concluded in January 2026 and received over 98,000 submissions, generating high performing AI models and novel computer vision approaches.

They represent state of the art performance and have been trained on a large, diverse dataset covering Australian pastures across seasons, regions, and species mixes.

AgTech companies, industry partners, and investors are now invited to collaborate on:

• Integrating winning models into commercial platforms
• On farm validation and pilot deployments
• Product development and go to market pathways
• Commercialisation and scale up

This is an opportunity to move proven AI capability from competition and research into practical, producer-ready tools.

Submit your EOI by enquiring now.

Opportunity background

CSIRO is Australia’s national science agency, and MLA is the research and marketing body for the Australian red meat sector. Together, they collaborated to advance pasture measurement capability through applied AI and data driven innovation.

The competition produced advanced AI and computer vision models capable of predicting pasture biomass at the species level, eliminating the traditional trade off between accuracy and scalability. These models also improve calibration and validation of satellite based approaches, strengthening integration with broader digital agriculture systems.

Accurate biomass prediction enables producers to meet residual pasture targets, avoid overgrazing and land degradation, and optimise regrowth. Precision grazing supported by AI improves long term pasture health, supports soil carbon retention, and reduces erosion risk. Economically, it enables better alignment of feed supply with livestock demand, reducing supplementary feeding, lowering operating costs, and reducing emissions intensity.

Potential other applications

The models could potentially be applied in: 

• Crop management – including yield estimation, nutrient management, irrigation optimisation, and weed or pest detection

• Forestry and carbon management – such as forest biomass estimation, deforestation monitoring, and reforestation planning; and 

• Environmental monitoring – covering grassland and rangeland management, biodiversity assessment, and soil erosion or land degradation monitoring.

  ---

  EVOKEAG17FEB2026

  IDEASTOIMPACT