What happened
Japan's Ministry of Agriculture, Forestry and Fisheries commissioned a project (JP17935051) in which NARO and prefectural institutes built a pest and disease diagnosis AI. It reads smartphone photos of tomato, cucumber, strawberry and eggplant. It was developed from more than 700,000 damage images collected by experts. Nihon Nohyaku built it into its free app レイミーのAI病害虫雑草診断 on 1 March 2022. The Kyoto Prefectural Agriculture, Forestry and Fisheries Technology Center collected 19,366 disease and 12,375 pest images of eggplant for training. It then had the AI diagnose newly photographed images, 2,613 of diseases and 2,363 of pests, and reports that accuracy reached 80% or more. A trial survey of 62 extension advisers and growers found expectations for the system alongside requests for higher accuracy. The Mie Prefecture Agricultural Research Institute, which also supplied training images, reports recall of roughly 80% or more and lists new farmers and trainees as the intended users. Both institutes took part in the project. No outside body checked the figures, and neither report describes any outcome for crops.
Sources
-
Report, by an evaluator that the makers or users commissioned.
京都府農林水産技術センター 生物資源研究センター 応用研究部 (Kyoto Prefectural Agriculture, Forestry and Fisheries Technology Center, Biological Resources Research Center, Applied Research Division) (2022). ＡＩを活用した画像によるナス病害虫診断技術の開発 (Development of an AI image-based eggplant pest and disease diagnosis technology).
AI の精度検証は、新たに撮影した病害虫被害画像を AI に診断させ、病害 2,613 件、虫害 2,363 件について実施し、最終的には 80%以上の精度を確保することができた。
-
Report, by an evaluator that the makers or users commissioned.
三重県農業研究所 (Mie Prefecture Agricultural Research Institute) (2022). 令和３年度 三重農研成果情報: スマートフォンカメラ撮影画像から病害虫を識別できる病害虫診断AI (FY2021 results: a pest and disease diagnosis AI that identifies pests and diseases from smartphone camera images).
病害虫診断AIによる診断精度は、概ね 80%以上の再現率です。
Details
AI system
- Workflow
-
The AI was developed from more than 700,000 pest and disease damage images collected by experts at agricultural research institutes. Kyoto's center photographed 19,366 disease and 12,375 pest images of eggplant by plant part, severity and time for training. The accuracy check used newly photographed images that were not part of training.
Follow-up
- Conditions for replication
-
The crop and the pest or disease must be among those the AI was trained on, ten for eggplant. The photo must be taken with a smartphone camera in the field, as in the checks.
Entry details
- Sector
-
Agriculture / crop pest and disease diagnosis.
- Purpose
-
NARO and prefectural institutes built the pest and disease diagnosis AI, under a MAFF project, to let growers without diagnostic experience identify pests and diseases from a smartphone photo.
31 more questions not yet checked
model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, AI contribution, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.
Cite this benefit
Peter H. Vartanian. (2026). Prefectures find 病害虫診断AI 80% accurate on new pest images (AIBD-1198). Artificial Intelligence Benefits Database, edition of 2 November 2026. https://petervartanian.xyz/aibd/records/AIBD-1198.html (archived at https://doi.org/10.7910/DVN/AADSII, version 9.0)