Synthetic Occlusion Augmentation with Volumetric Heatmaps for the 2018 ECCV PoseTrack Challenge on 3D Human Pose Estimation
arXiv:1809.04987
Abstract
In this paper we present our winning entry at the 2018 ECCV PoseTrack Challenge on 3D human pose estimation. Using a fully-convolutional backbone architecture, we obtain volumetric heatmaps per body joint, which we convert to coordinates using soft-argmax. Absolute person center depth is estimated by a 1D heatmap prediction head. The coordinates are back-projected to 3D camera space, where we minimize the L1 loss. Key to our good results is the training data augmentation with randomly placed occluders from the Pascal VOC dataset. In addition to reaching first place in the Challenge, our method also surpasses the state-of-the-art on the full Human3.6M benchmark among methods that use no additional pose datasets in training. Code for applying synthetic occlusions is availabe at https://github.com/isarandi/synthetic-occlusion.
arXiv commentExtended abstract for the 2018 ECCV PoseTrack Workshop, updated with full result tables
References in corpus (7)
- YOLOv3: An Incremental Improvement
- Improved Regularization of Convolutional Neural Networks with Cutout
- VNect: Real-time 3D Human Pose Estimation with a Single RGB Camera
- Random Erasing Data Augmentation
- Numerical Coordinate Regression with Convolutional Neural Networks
- How Robust is 3D Human Pose Estimation to Occlusion?
- 3D Human Pose Estimation with 2D Marginal Heatmaps
Cited by in corpus (9)
- Vision-Based Human Pose Estimation via Deep Learning: A Survey
- Anti-UAV: A Large Multi-Modal Benchmark for UAV Tracking
- SMPLR: Deep SMPL reverse for 3D human pose and shape recovery
- Advancing 6-DoF Instrument Pose Estimation in Variable X-Ray Imaging Geometries
- Metric-Scale Truncation-Robust Heatmaps for 3D Human Pose Estimation
- View-Invariant, Occlusion-Robust Probabilistic Embedding for Human Pose
- DeepFuse: An IMU-Aware Network for Real-Time 3D Human Pose Estimation from Multi-View Image
- Learning to Generate Synthetic Data via Compositing
- Explicit Spatiotemporal Joint Relation Learning for Tracking Human Pose