Pocket Agronomist

Pocket Agronomist is a crop scouting application designed to support growers, agronomists, field interns, and other agricultural professionals tasked with monitoring crop development and field conditions. The application uses augmented reality and convolutional neural networks to perform two common crop scouting tasks: stand counts and disease identification.

Unlike other crop scouting applications requiring manual data entry, form fill-outs, checklists or hierarchy trees, Pocket Agronomist automates the stand counting and disease identification processes, allowing you to simply point your iPhone or iPad camera at the crop and begin receiving insights.

In place of time-consuming manual measurements for plant distance, Pocket Agronomist makes gathering stand count statistics easy with the aid of augmented reality. You simply aim your camera at the ground near a plant and tap to locate it. As each plant is identified in a row, their real-world locations are tracked, shown in a live 3-D overlay on the camera feed, and the distances between them precisely measured. When done, the application automatically calculates the following stand count metrics: the total number of plants within the count, the average distance between each plant, the standard deviation of the count, the estimated plant population based on the count, and an estimated yield loss resulting from uneven spacing. These results can then be averaged with other stand counts from within the same field or sent to the appropriate parties for further reporting.

In addition to stand counts, Pocket Agronomist utilizes high-performance convolutional neural networks to diagnose and localize crop diseases from live camera video. Using leading edge object detection techniques, crop diseases are detected and diagnosed in real time when you point your device at a plant. You don't have to line up for a perfect photo or upload anything to a server, a labeled box will be drawn around any disease detected in the live camera video. All analysis is performed on device and in real time, allowing the application to function in regions without data connectivity (a common situation with remote farm fields).

Currently, Pocket Agronomist is trained to identify 12 foliar corn diseases and man-made anomalies prevalent in the United States, including: Anthracnose Leaf Blight, Bacterial Leaf Streak, Common Rust, Eyespot, Goss’s Bacterial Wilt, Gray Leaf Spot, Northern Corn Leaf Blight, Northern Leaf Spot, Physoderma Brown Spot, Common Smut, Southern Rust and Urea Burn.

The disease diagnosis functionality is under active development, so disease accuracies may vary. Do not solely rely on the identifications provided and always seek out the recommendations of a trained agronomist.

We also would like your help to improve the existing diagnostic cases and to expand the breadth of crops supported by the application. If you opt to do so, images you capture using the diagnostic functionality will be anonymized and used to better train our diagnostic neural networks. Pocket Agronomist also has a dedicated image capture tab that can be used to collect and send us examples of plants and diseases you would like to see added to the application.