Žan Ambrožič, Lorenzo Bianco, Rok Šturm, David Susič, Maj Smerkol and Anton Gradišek

Abstract
Passive sensing of pollinator activity is important for biodiversity
monitoring and conservation, yet conventional acoustic or visual
methods produce large amounts of data and face deployment
challenges. In this work, we present initial results on investigating
stem vibration as an alternative signal for detecting pollinator
presence on flowers. Vibration recordings were collected with
a laser vibration instrument from various flower species at multiple
locations in Slovenia, totaling approximately 14 hours, of
which 3 hours were expert-annotated for insect activity. The
task was formulated as a binary classification problem: determining
whether a vibration segment corresponds to a pollinator
physically touching the flower. Using a neural network model,
performance was evaluated with five-fold cross-validation across
three experiments: (i) using a balanced subset, (ii) using the full
dataset, and (iii) using the full dataset with heuristic prediction
smoothing. On the balanced subset, the model achieved an average
F1-score of 0.86 ± 0.06; on the full dataset, 0.62 ± 0.07;
and with heuristic smoothing, 0.69 ± 0.11, demonstrating both
the feasibility of vibration-based detection and the benefits of
post-processing. Beyond binary detection, future work will focus
on species- and activity-level classification. Ultimately, the goal
is to develop lightweight vibration detectors deployable directly
on plants, enabling scalable estimation of pollinator visitation
rates in natural and agricultural environments.