Case Study

Teaching AI to See What Farmers See

Herd-i partnered with IndiVillage for expert video annotation services that support AI-powered livestock health detection and market expansion.

HERD-I

Lameness and ill health in cattle globally result in an annual loss of approximately $6 billion for the farming sector. This issue is particularly challenging because sick cows instinctively conceal their condition by hiding within the herd.

Herd-i is a New Zealand-based dairy technology company developing camera systems for early lameness detection and body condition scoring in cattle. As the company expanded into the US market - adapting its models for Holstein breeds and concrete floor environments - it needed a data annotation partner that could deliver precision, consistency, and scale. IndiVillage became that partner, providing expert video annotation services that underpin Herd-i's AI-powered livestock health platform.

10,000+

video clips processed weekly

3-day

turnaround SLA consistently met

Challenge

Teaching AI to See What Farmers See

Herd-i's camera systems capture video of cows as they walk, but turning that footage into reliable health assessments requires vast quantities of precisely labelled training data. Each video clip demands accurate identification of individual animals - matching electronic identification (EID) tags to specific timestamps - under conditions that include variable lighting, black-coated cattle, and numbers that fluctuate as animals move through the frame.

Previous annotation providers had struggled with consistency. Herd-i needed a team that could handle the nuance of agricultural video data at scale, deliver within tight turnaround windows to support continuous model improvement, and operate under a reliable commercial framework. With plans to onboard four new farm customers per week, the annotation pipeline had to scale in lockstep with the business.

Solution

Dedicated Agricultural Annotation Expertise

IndiVillage deployed a dedicated team with deep experience in agricultural data labelling. Working within Herd-i's proprietary annotation portal, the team performed timestamp-based cow identification matching - confirming EID numbers at specific moments in each video, flagging unstable readings, and handling edge cases such as delayed drafting caused by timing complications with Information provided by an upstream supplier technology.

Quality was maintained through rigorous internal review processes, with every annotation batch subject to verification before delivery.

Results

Reliable Data Fuelling Market Expansion

IndiVillage's consistent, high-quality annotation output has enabled Herd-i to accelerate its model development across both lameness detection and body condition scoring. Multiple annotation projects have been delivered successfully - from the initial proof of concept with 200-300 cow samples through to ongoing production batches - each meeting the agreed SLA and accuracy standards.

With 96% staff retention on the dedicated team, IndiVillage provides the continuity and domain familiarity that complex agricultural annotation demands. The relationship continues to deepen, with discussions underway around broader AI development support as Herd-i's platform evolves.