Crop-row weeding navigation capture

This is a custom capture program for row-crop weeding navigation between plants. Operators or supervised autonomy drive your implement down real rows while we record RGB or RGB-D, vehicle state, and plant-versus-weed labels, including missed plants and lodging. It exists because field lighting, dust, and crop stage change the discrimination problem in ways a generic outdoor locomotion log does not cover.

What we collect

Row-tracking episodes with plant discrimination labels across crop stage, spacing, and lighting. Missed plants, lodged stems, and glare washouts are sampled on purpose so evaluation can slice navigation from weeding accuracy.

How capture works

Scoping locks crop, growth stage, implement width, and the plant/weed taxonomy. A short pilot checks camera height and label quality, then capture scales across rows using our robotics data collection process. Dust and lodging tails follow edge-case data collection.

Who it is for

Agricultural robotics teams whose weeder must stay in-row without damaging crop. The program sits under agricultural robotics data collection and is distinct from orchard fruit-pick teleoperation, which is a canopy manipulation problem rather than row navigation.