Deep Multi-Task Networks For Occluded Pedestrian Pose Estimation
This work addresses a critical gap in automotive safety by enabling pose estimation for occluded pedestrians, which is incremental as it builds on existing multi-task and domain adaptation methods.
The paper tackles the problem of estimating pedestrian poses under occlusion by proposing a multi-task framework that combines detection and instance segmentation from automotive and non-automotive datasets, using unsupervised domain adaptation to learn pose-specific features, resulting in improved state-of-the-art performances in pose estimation, pedestrian detection, and instance segmentation.
Most of the existing works on pedestrian pose estimation do not consider estimating the pose of an occluded pedestrian, as the annotations of the occluded parts are not available in relevant automotive datasets. For example, CityPersons, a well-known dataset for pedestrian detection in automotive scenes does not provide pose annotations, whereas MS-COCO, a non-automotive dataset, contains human pose estimation. In this work, we propose a multi-task framework to extract pedestrian features through detection and instance segmentation tasks performed separately on these two distributions. Thereafter, an encoder learns pose specific features using an unsupervised instance-level domain adaptation method for the pedestrian instances from both distributions. The proposed framework has improved state-of-the-art performances of pose estimation, pedestrian detection, and instance segmentation.