PineSORT + AinnovaTech — Costa Rica pineapple plant-counting AI cluster (drone-based CV)

Central-America (Costa Rica canton Río Cuarto, Alajuela province; Pineapple Cámara Nacional)

Content

The Costa Rica pineapple AI cluster is the corpus’s first commodity-region cluster unit — two distinct computer-vision deployments at the same crop (pineapple) and the same region (Costa Rica), surfacing both academic-research-led and commercial-CRC-led deployment pathways for the same agronomic question (counting plants at the flowering / early-growth stages).

PineSORT (academic, drone-based): Drone-based system for counting pineapple plants in open-field production, presented by Fabián Fallas Moya (Professor and researcher at UCR — Universidad de Costa Rica — and TEC — Costa Rica Institute of Technology) at the IICA Digital Agriculture Week 2025 AI Day (San José, Costa Rica, October 2025). Origin research: Andrés Castillo Leiton, Agrointeligencia Geoespacial, working with La Cámara de Piñeros Unidos de Costa Rica (Committee of United Pineapple Sowers of Costa Rica). The research project, named “State-of-the-art Convolutional Neural Network applied in the automatic detection of pineapple plants from RPAS images”, was conducted at one of the cooperative’s properties in the cantón de Río Cuarto, Alajuela province, on young plants at least two weeks old.

Documented scope (per SPH Engineering case study, August 2022):

• Manual baseline: 2 to 2.5 hectares counted per day (1 person with mechanical counter, “María”)
• Drone-equipped baseline: 50 hectares counted per day (DJI Phantom 4 Pro with 1” 20-megapixel sensor, flight height 45 m, ground sampling distance 1.6-2 cm/pixel, 85% horizontal and vertical overlap, UgCS flight planning)
• 20× productivity gain confirmed in research deployment
• Processing pipeline (SPH Engineering Consulting & Development): training image set preparation → object detection model training (CNN) → algorithm for field-boundary-separated result aggregation
• End-to-end processing cycle: 24 hours from image capture to final report

AinnovaTech (commercial, vision-based flowering counting): Computer-vision solution for automated counting during the flowering stage of pineapple growth, presented by Rodrigo Herrera Garro, co-founder and CTO of AinnovaTech, at the IICA DAW 2025 AI Day. The product is positioned for AI “to connect worlds: medicine, agriculture, engineering… Farmers are open to trying, they don’t close the door, and that creates opportunities” (Herrera Garro quote, IICA Oct 2025).

Why a cluster unit, not two separate units:

PineSORT and AinnovaTech are structurally distinct:

• Different actor-types (academic-research-led cooperative project vs commercial-CRC startup)
• Different growth-stage application (young-plant counting vs flowering-stage counting)
• Different deployment status (academic pilot at named cooperative property vs commercial; specific named customer roster not surfaced in primary sources)
• Different vantage-point decisions (drone-equipped at scale vs flowering-stage-specific CV)

But they share:

• Same crop (pineapple)
• Same region (Costa Rica, with IICA convening role)
• Same institutional-event platform (DAW 2025 AI Day)
• Same analytical problem (plant counting / yield forecasting)

Treating them as two separate units would lose the cluster-pattern observation: Costa Rica has co-incident academic and commercial pineapple-AI development in 2025, with the IICA institutional convening role as the surfacing platform. The cluster unit preserves the observation without forcing it into a single vendor or a single deployment.

Strategic significance for the corpus:

1. Pineapple is one of Costa Rica’s signature crops. Costa Rica is the world’s largest pineapple exporter; the Corporación Bananera Nacional-era agricultural data infrastructure is concentrated on pineapple and banana production for export. AI deployment at this commodity touches the country’s food-economy centre of gravity.
2. The Cámara de Piñeros Unidos (Pineapple Producers Union of Costa Rica) is a farmer-led institutional anchor — distinct from government-led or corporate-led AI deployment models. AI deployment at a farmer-cooperative property is a structurally-distinct pathway worth surfacing.
3. IICA DAW 2025 as the institutional convening platform — the unit’s institutional context is the IICA-organised AI Day, not a vendor-led or government-led event. This is the corpus’s first academic-research-deployment unit with a strong regional-institutional convening role documented.
4. 20× productivity gain is concrete and reproducible — drone-equipped pineapple counting at 50 ha/day vs manual at 2.5 ha/day is the corpus’s most concrete named-productivity-GAIN number for crop-counting AI in open-field production, with named deployment country, named cooperative property, named researcher.
5. Object-detection CNN for pineapples specifically — the corpus’s computer-vision taxonomy cell for on-farm open-field AI now has a Costa Rica entry, complementing NA / EU / India entries.

What this unit is doing in the taxonomy

Anchors the commodity-region cluster pattern for on-farm open-field AI, complementing:

• bayer-climate-fieldview.md (NA multi-continent, multi-crop)
• agrosmart-brazil.md (Brazil multi-crop smallholder venture-platform)
• wagri-japan-agricultural-data-platform.md (Japan state-DPI substrate, multi-crop)
• india-digital-agriculture-mission-agristack.md (India state DPI, multi-crop)
• kilimo-argentina-irrigation.md (Argentina irrigation AI, multi-crop)
• auravant-argentina-precision-agriculture.md (Argentine SaaS agronomy)

Structurally distinct from the deployment model spectrum (cloud-mediated / vertically integrated / cooperative-mediated / venture-funded-platform / state-affiliated-industrial). PineSORT sits closer to cooperative-mediated + academic-anchored. AinnovaTech sits closer to commercial-CRC-anchored. Both share the commodity-region cluster attribute.

Why it matters for talks

• The concrete 20× productivity number (50 ha/day drone vs 2.5 ha/day manual) is the corpus’s most quotable single-deployment statistic for crop-counting AI in open-field production; useful for talks that need hard productivity numbers.
• The Cámara de Piñeros Unidos anchor is the corpus’s clearest LAC farmer-cooperative AI deployment with named property — useful for the deployment-model spectrum frame.
• The DAW 2025 institutional convening role demonstrates how IICA’s convening-power surfaces cluster patterns that vendor-listing searches miss. Useful methodological lesson: institutional-event proceedings are primary sources for LAC deployments, distinct from the NA pattern where vendor customer-story pages are primary.
• PineSORT’s CNN architecture (state-of-the-art Convolutional Neural Network) — substantive technical content; useful for talks that need to specify architecture type rather than vague “AI/ML”.

Critical context

• AinnovaTech deployment scope is not surfaced in primary sources. The IICA DAW 2025 announcement names the technology and the founders, but does not name a specific deployment customer or a deployment scale. G-086 candidate for follow-up verification.
• PineSORT is academic-research scope, not commercial deployment. The 20× productivity gain is from the named research project at the named cooperative property; commercial deployment at scale (e.g. multiple cooperative properties under commercial contract) is not surfaced. G-094 candidate.
• DJI Phantom 4 Pro dependence. PineSORT’s drone component requires the DJI drone ecosystem — a dependency on Chinese-manufactured hardware in a small Costa Rican deployment context. The corpus’s broader framing of DJI Agriculture as a multi-continent vendor (in units/dji-agriculture-global-export.md) makes this not surprising, but worth flagging for the regional cluster pattern: Costa Rica AI deployments rely partly on Chinese-manufactured drone hardware, analogous to how Japanese AI deployments rely on Yamaha drone hardware (see yamaha-fazer-agricopter-drone-japan.md).
• Connectivity infrastructure (a 13% LAC broadband access, rural-urban gap 25%) limits the data-upload side of drone-based CV deployment at smallholder-cooperative scale.
• Farmer-led institutional anchor. The Cámara de Piñeros Unidos is farmer-led, not state-led. The Indonesian / Vietnamese equivalent patterns likely share this farmer-cooperative deployment pathway. Worth tracking for an Indonesia / SEA cycle later.
• AI Day framing caveat. IICA DAW 2025 framed this as “Innovating Agriculture with AI: From Vision to Action” — explicitly noting that “AI is no longer the future; it is the tool of the present” with deployment-specific examples. The surfacing role of IICA is documented but the deployment scale is named only in the PineSORT case.
• Fabián Fallas Moya dual institutional role (UCR + TEC) reflects Costa Rica’s inter-university research collaboration pattern. Distinct from NA NSF/USDA-NIFA multi-state Land-grant collaboration model.