Event-Based and Frame-Based Dataset of Deformable Linear Objects

The "Event-Based and Frame-Based Dataset of Deformable Linear Objects" is a collection of recordings of deformable linear objects (DLOs) under robotic manipulation. The dataset is designed to serve as a benchmark for DLO tracking algorithms, with a particular focus on scenarios involving fast motion, motion blur, multi-instance occlusion, and identity preservation across long manipulation sequences. It provides an valuable resource for researchers and practitioners in the fields of robotics, computer vision, and industrial automation. A unique aspect of this dataset is the simultaneous capture from two complementary sensor modalities: a high-resolution event camera (Prophesee EVK4 HD, 1280×720) and a frame camera (IDS U3-3680XCP, 2592×1944). The subset of single DLO sequences additionally includes recordings from an Intel RealSense D415 RGB-D camera, enabling direct comparison with depth-sensing tracking methods. This pairing enables research into hybrid tracking approaches that combine the high temporal resolution of event cameras with the spatial richness of conventional frame-based imaging. The dataset covers three industrially representative DLO types manipulated by a six-axis industrial robot at three speed levels. Two experimental scenarios are included: a multi-DLO scene with four overlapping cables for identity tracking and ID-switch evaluation, and a single-DLO scene for quantitative tracking accuracy benchmarking. In total, the dataset comprises measurement sequences with 2,300 frames and 5.5 GB of event data, accompanied by per-frame ground-truth polyline annotations and calibration files. The ground-truth annotations and sensor calibration data make this dataset well-suited for training and evaluating tracking, segmentation, and state estimation algorithms operating on event streams, frames, or both. It enables advancing real-time DLO tracking methods applicable to automated cable routing, robotic assembly, and quality control in industrial manufacturing environments.