Detecting Subacute Ruminal Acidosis with a Real-Time Deep Learning Algorithm
A precision-livestock platform that reads cattle rumen gas emissions through optical gas imaging — turning thermal video into an early, non-invasive signal of digestive health.
A costly disorder that hides in plain sight.
Subacute ruminal acidosis (SARA) is one of the most common and economically damaging metabolic disorders in dairy and beef cattle. It depresses feed efficiency, milk yield, and animal welfare — yet it produces few outward clinical signs, making it notoriously difficult to detect on the farm.
Conventional diagnosis relies on invasive rumen pH sampling. Our project asks a different question: what if the rumen's own gas emissions could reveal its health — captured at a distance, in real time, with a camera?
From a breath of gas to a health signal — in four steps.
Accurate, lightweight, and fast
enough for the field.
Three papers, one growing system.
CarboFormer
A lightweight semantic segmentation architecture for CO₂ detection using optical gas imaging.
FUME
A fused, unified multi-gas emission network jointly segmenting CO₂ and methane from livestock rumen.
VLMDual
A vision-language distillation framework for joint classification and plume segmentation in SARA detection.
From controlled chambers to live cattle.
Sharing the work beyond the lab.
This work is supported by the United States Department of Agriculture, National Institute of Food and Agriculture (USDA-NIFA), through the Capacity Building Grants for Non-Land-Grant Colleges of Agriculture.