Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

YCB-Ev SD: Synthetic event-vision dataset for 6DoF object pose estimation

Pavel Rojtberg, Julius Kühn

TL;DR

This work tackles the scarcity of synthetic event-based data for 6DoF object pose estimation by introducing YCB-Ev SD, a large SD-resolution synthetic event dataset derived from PBR YCB-Video scenes with simulated linear camera motion. It systematically compares event representations for CNN-based pose inference, showing that time-surfaces with linear decay and dual-channel polarity encoding yield the best performance, with polarity providing the strongest gains and linear temporal encoding preserving motion information. The dataset is provided in a BOP-aligned structure including raw event streams and precomputed optimal histograms to support immediate benchmarking and reproducibility. Overall, the work enables robust evaluation of event-based pose estimation methods and motivates future real-world validation, richer motion patterns, and transformer-based approaches.

Abstract

We introduce YCB-Ev SD, a synthetic dataset of event-camera data at standard definition (SD) resolution for 6DoF object pose estimation. While synthetic data has become fundamental in frame-based computer vision, event-based vision lacks comparable comprehensive resources. Addressing this gap, we present 50,000 event sequences of 34 ms duration each, synthesized from Physically Based Rendering (PBR) scenes of YCB-Video objects following the Benchmark for 6D Object Pose (BOP) methodology. Our generation framework employs simulated linear camera motion to ensure complete scene coverage, including background activity. Through systematic evaluation of event representations for CNN-based inference, we demonstrate that time-surfaces with linear decay and dual-channel polarity encoding achieve superior pose estimation performance, outperforming exponential decay and single-channel alternatives by significant margins. Our analysis reveals that polarity information contributes most substantially to performance gains, while linear temporal encoding preserves critical motion information more effectively than exponential decay. The dataset is provided in a structured format with both raw event streams and precomputed optimal representations to facilitate immediate research use and reproducible benchmarking. The dataset is publicly available at https://huggingface.co/datasets/paroj/ycbev_sd.

YCB-Ev SD: Synthetic event-vision dataset for 6DoF object pose estimation

TL;DR

This work tackles the scarcity of synthetic event-based data for 6DoF object pose estimation by introducing YCB-Ev SD, a large SD-resolution synthetic event dataset derived from PBR YCB-Video scenes with simulated linear camera motion. It systematically compares event representations for CNN-based pose inference, showing that time-surfaces with linear decay and dual-channel polarity encoding yield the best performance, with polarity providing the strongest gains and linear temporal encoding preserving motion information. The dataset is provided in a BOP-aligned structure including raw event streams and precomputed optimal histograms to support immediate benchmarking and reproducibility. Overall, the work enables robust evaluation of event-based pose estimation methods and motivates future real-world validation, richer motion patterns, and transformer-based approaches.

Abstract

We introduce YCB-Ev SD, a synthetic dataset of event-camera data at standard definition (SD) resolution for 6DoF object pose estimation. While synthetic data has become fundamental in frame-based computer vision, event-based vision lacks comparable comprehensive resources. Addressing this gap, we present 50,000 event sequences of 34 ms duration each, synthesized from Physically Based Rendering (PBR) scenes of YCB-Video objects following the Benchmark for 6D Object Pose (BOP) methodology. Our generation framework employs simulated linear camera motion to ensure complete scene coverage, including background activity. Through systematic evaluation of event representations for CNN-based inference, we demonstrate that time-surfaces with linear decay and dual-channel polarity encoding achieve superior pose estimation performance, outperforming exponential decay and single-channel alternatives by significant margins. Our analysis reveals that polarity information contributes most substantially to performance gains, while linear temporal encoding preserves critical motion information more effectively than exponential decay. The dataset is provided in a structured format with both raw event streams and precomputed optimal representations to facilitate immediate research use and reproducible benchmarking. The dataset is publicly available at https://huggingface.co/datasets/paroj/ycbev_sd.

Paper Structure

This paper contains 5 sections, 2 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction and related work
  2. Synthetic event-data generation
  3. Comparing event representations for CNN inference
  4. Dataset structure and usage
  5. Conclusion and future work

Figures (3)

  • Figure 1: Sythethic RGB renderings (left), matching event histogram (middle), and depth image (right) from the our dataset. Throughout this paper, the polarity of the events is represented as blue (falling) and green (rising).
  • Figure 2: Common event representations: (left) event histogram and (right) time-surface
  • Figure 3: Time-surface representations with polarity encoding: (left) 2 channels in blue-green (right) 1 channel (black = off, white = on)