TaranisSame team, four growing seasons, 460+ weed species held in taxonomy.
4.5M+ images · 460+ species · 99.4% sustained
Read case →02 · The challenge
Millions of pest images from different regions, different crop cycles, different growth stages. Models trained on inconsistent annotation drift in the field — at production volume that drift is invisible until it costs growers.
FMC’s Arc Farm Intelligence required precise pest detection and species classification across diverse agricultural environments. The bottleneck was annotation: millions of images requiring taxonomy consistency and research-grade accuracy before any model could deploy.
Without a reliable annotation backbone, Arc’s models would fail in the field. Pests mis-identified as harmless cost growers money and trust. Previous workflows could not hold biological taxonomy at scale.
03 · How we did it
Precision annotation designed for agricultural computer vision.
01
Expert taxonomy
We aligned annotation protocols with FMC’s biological classification standards before a single image was tagged. Each pest species received a defined, verifiable taxonomy. This reduced downstream model confusion and meant Arc could return actionable species-level predictions to growers.
7 production AI models
02
Precise geometry
Every pest instance received bounding box labelling to sub-pixel precision. Models trained on sloppy boxes fail in the field. Our annotators were trained on Arc’s accuracy standard, not generic labelling rules.
96%+ precision achieved
03
Consistency at scale
The same core team maintained tagging protocols across hundreds of thousands of images. Retention compounds knowledge — our 96% annual staff retention means the annotators building Arc’s dataset in month 3 are still refining it in month 12. Consistency accumulates.
16 years same discipline
04
Human–AI benchmarking
We embedded AI-assisted review tools into the workflow, allowing our team to catch edge cases and train Arc’s models simultaneously. Hybrid throughput without sacrificing the precision that production deployment demands.
300K+ hectares covered
04 · The outcome
FMC’s Arc platform now covers 300,000+ hectares globally. The 96%+ precision on FMC’s seven production models means growers receive accurate, species-level pest intelligence. That accuracy came from a data foundation where every bounding box, every taxonomy decision, and every edge case was deliberate.
Quality compounds. Consistent annotation over years builds domain expertise that shows in the model. Farmers benefit with actionable insights that reduce chemical use and improve yields. Arc’s competitive edge rests on data rigour that cascades through every deployment.
05 · In their words
“IndiVillage’s precise image tagging enables us to optimize workflows and deliver value to growers worldwide.”
Joseph Davis · Director of Data Science & Data Platforms · FMC
06 · The numbers underneath
300K+
Hectares
covered
7
AI models
in production
96%+
Pest detection
precision
07 · Other programmes that shipped
Same operating discipline. Different modalities.
08 · Work with us
Run an annotation audit on your data.
Send us 100 frames in any modality — image, video, LiDAR, audio, text. We'll return annotated output, an accuracy benchmark, and a programme recommendation in 48 hours.