Multi-stage vision system improves crop disease and pest diagnosis accuracy
Configurable Multi-Stage Vision Pipeline for Crop Disease and Pest Diagnosis
Computer Vision and Pattern RecognitionComputation and LanguageMachine Learning
Summary
Farmers often send photos of their crops when they see something wrong, but the photos can be blurry, dark, or unclear, making diagnosis hard. The authors studied over a million images and found many were rejected or incorrectly labeled. They created a system with three steps: checking photo quality, identifying the crop, and then finding the disease or pest. This new multi-stage system is more accurate and flexible than the current one, handling poor images better and always giving an answer.
What this means in practice
- •For agriculture technology developers: Develop mobile advisory apps that reliably diagnose crop issues from low-quality farmer photos using a multi-stage model pipeline.$Commercial implications: Enables creation of sellable crop diagnostic services for smallholder farmers using flexible, accurate AI models adapted to resource-constrained settings.
- •For machine learning engineers: Implement modular vision systems with separate quality gating, crop detection, and disease diagnosis stages for better model control and accuracy.
Authors
Naga Ganesh, Chandrashekar M S, Lakshmi Pedapudi, Aakash Singh, Vineet Singh
Abstract
Farmer.Chat is Digital Green's farm advisory service for smallholder farmers. When something looks wrong with a crop, the farmer takes a photograph and sends it, and that photograph is the whole question: no symptom described, no crop named, often no text at all. The service has to determine whether the picture can be used, what crop it shows, and what is wrong with it, from images taken on cheap phones in a field, in poor light and with a moving camera. The system doing this today cannot be adjusted. It has no adjustable thresholds for photograph rejection, crops and problems cannot be added, and there is no confidence cut-off to set. We study about 1.16 million photographs sent to Farmer.Chat from Ethiopia, India, Kenya and Nigeria. The production quality gate rejected 46.8% of the images it judged, over a quarter of those reaching diagnosis returned no crop name, and 35.8% of the labelled problems filed under "disease" are pests, identifiable without the crop. We therefore split the work into three stages: a quality gate (M0), a crop detector (M1), and a disease or pest detector (M2). Route A fills all three with one fine-tuned vision-language model (Qwen3-VL-4B) answering in a single call. Route B fills each with a small specialist model (DaViT, YOLO26). We replace our production GPT-4o quality gate with a small MobileNetV3 gate at 86.9% F1 in 12 ms. On one test set scored the same way for every system, a hierarchical DaViT-Base achieves 95.41% crop accuracy against 91.46% for the production baseline. It also leads on diagnosis and never declines to answer, while every language model in the comparison leaves a large share of rows with no diagnosis. The fine-tuned model retains two capabilities the specialists do not have: one call for all three stages, and a request for a better photograph when the image cannot support an answer.