Validate livestock sensors against video
Line up collar or ear-tag accelerometer data with video, label behaviour once, and measure how well a classifier or a commercial sensor matches what the animals actually did.

Finished numbers are not enough
Collars, ear tags and leg sensors report grazing, rumination, lying time or activity scores, worked out by algorithms you usually can't inspect. To check those numbers, or to build and test your own classifier, you need the raw sensor data labelled against what the animals actually did, mostly from video. That alignment and labelling is where most of the time goes: video players, annotation tools and merge scripts that don't talk to each other.
Firetail puts sensor data, video, labels and predictions on one timeline. It comes from wildlife research, where tags record GPS and acceleration in bursts, and the workflow is the same for cattle, sheep and goats on pasture or in the barn.
The workflow
- Open the sensor data: GPS, acceleration, gyroscope and magnetometer, via Movebank or from files (see
- supported tags and file formats). Continuous recordings are cut into
- bursts.
- Synchronize the video: replay footage in step with the data and align both on an event you can see in each
- (see Synchronize video with accelerometer data). One video
- can serve several animals with their own offsets.
- Label behaviour: annotate directly in Firetail, or import events from BORIS (tabular CSV exports from BORIS v7
- and v8, including point and range events and modifiers).
- Classify: let FireSOM find behaviour categories in the acceleration data (see
- Classify animal behaviour from accelerometer data), or import
- predictions from your own model or from a sensor's output with Firetail's annotation exchange format.
- Evaluate: choose a reference (your video labels) and a prediction (a model, a sensor, a second observer), and
- Firetail reports precision, recall and F-measure per behaviour, computed from overlapping segments. The same
- comparison works between two human observers.
Your loggers
Firetail reads the tags and formats listed under supported tags and file formats, and CSV files with timestamped acceleration. Data from generic loggers or vendor exports that aren't on the list: we offer conversion as a service, and support for your format may well come in the next release.
Free one-hour consultation
Tell us about your study, your sensors and video, and your questions. In a one-hour online meeting we look at them together and see whether Firetail fits. Get in touch.
Get Firetail
free downloadGet started with Firetail today! Interactively process high-density data for multi-animal, multi-year projects!
“No risk: request a three-month trial, kickstart your analysis and see your data in an all-new light!”
