Insect Camera Trap Motion-Based Compression and Tracking Dataset

High-resolution camera trap video enables detailed study of insect behaviour, including fine-scale movement and flower visitation patterns, yet its storage, transmission, and processing demands present substantial challenges for field-deployed edge devices. We introduce a novel end-to-end system that combines a motion analysis–based video compression algorithm, purpose-built for camera trap applications, with a customised video processing workflow for automated extraction of behavioural data from compressed footage. The system is evaluated through a case study on insect–pollinator motion tracking, benchmarked across three widely used edge computing platforms.
The dataset provides the complete outputs required to reproduce and extend this evaluation. It includes processed videos with overlaid trajectories, raw and post-processed insect tracking CSV files, flower detection and tracking data, configuration files, and the pre-trained YOLO-based object detection models used in the pipeline. It also contains quantitative performance summaries used to assess compression efficiency and computational cost, including per-frame pixel-change statistics with full intensity-difference histograms, energy consumption measurements (three replicates per platform) for Jetson Nano, Raspberry Pi 4, and Raspberry Pi 5 devices, and detailed runtime logs from repeated processing trials across datasets and hardware platforms.