Establish the minimum resolution for aerial SPFES assessment

Establish the minimum image resolution at which Sheep Pain Facial Expression Scale cues remain recoverable for aerial imagery, in order to determine whether facial pain assessment can be deployed from unmanned aerial vehicles.

Background

The paper evaluates sheep facial pain recognition using close-range, ground-level images, where facial action units and SPFES cues can be observed. It discusses unmanned aerial vehicles as a potential component of a broader livestock-monitoring pipeline but distinguishes animal detection and localization from close-range facial pain assessment.

Because aerial platforms operate at greater distances, their imagery may not preserve the spatial detail required to recover clinically meaningful facial cues. The authors explicitly state that the system has not been tested on aerial images and identify the minimum recoverable resolution as a prerequisite for making claims about aerial pain scoring.

References

Our images are taken from the ground at close range. Facial action units need a spatial resolution that is not available at normal UAV flight altitudes, and we have not tested any part of this system on aerial images. A realistic role for a UAV in this pipeline is to detect and locate animals that deserve closer inspection . The facial pain assessment would then be carried out by a sensor at ground level or at low altitude. Before any claim about aerial pain scoring can be made, the minimum resolution at which SPFES cues remain recoverable must be established. We regard this as future work rather than a demonstrated capability.

When Do Language-Grounded Explanations Help? A Graph-Bottleneck for Farm Monitoring Interpretable Sheep Facial Pain  (2609.20427 - Noor et al., 17 Sep 2026) in Section 6, Discussion, subsection “Efficiency and the path to field deployment (LoRA)”

Finally, the images are taken from the ground at close range and come from a single annotated corpus. Generalization across breeds, housing systems, lighting conditions, and acquisition platforms remains untested, and the issue applies in particular to aerial imagery.