Moira Shooter, Charles Malleson, Adrian Hilton
Estimating the pose of animals can facilitate the understanding of animal motion which is fundamental in disciplines such as biomechanics, neuroscience, ethology, robotics and the entertainment industry. Human pose estimation models have achieved high performance due to the huge amount of training data available. Achieving the same results for animal pose estimation is challenging due to the lack of animal pose datasets. To address this problem we introduce SyDog: a synthetic dataset of dogs containing ground truth pose and bounding box coordinates which was generated using the game engine, Unity. We demonstrate that pose estimation models trained on SyDog achieve better performance than models trained purely on real data and significantly reduce the need for the labour intensive labelling of images. We release the SyDog dataset as a training and evaluation benchmark for research in animal motion.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Pose Estimation | StanfordExtra | PCK@0.1 | 78.65 | 8 Stacked Hourglass Network |
| Pose Estimation | StanfordExtra | PCK@0.1 | 77.19 | 2 Stacked Hourglass Network |
| Pose Estimation | StanfordExtra | PCK@0.1 | 50.77 | Mask R-CNN |
| 3D | StanfordExtra | PCK@0.1 | 78.65 | 8 Stacked Hourglass Network |
| 3D | StanfordExtra | PCK@0.1 | 77.19 | 2 Stacked Hourglass Network |
| 3D | StanfordExtra | PCK@0.1 | 50.77 | Mask R-CNN |
| Animal Pose Estimation | StanfordExtra | PCK@0.1 | 78.65 | 8 Stacked Hourglass Network |
| Animal Pose Estimation | StanfordExtra | PCK@0.1 | 77.19 | 2 Stacked Hourglass Network |
| Animal Pose Estimation | StanfordExtra | PCK@0.1 | 50.77 | Mask R-CNN |
| 1 Image, 2*2 Stitchi | StanfordExtra | PCK@0.1 | 78.65 | 8 Stacked Hourglass Network |
| 1 Image, 2*2 Stitchi | StanfordExtra | PCK@0.1 | 77.19 | 2 Stacked Hourglass Network |
| 1 Image, 2*2 Stitchi | StanfordExtra | PCK@0.1 | 50.77 | Mask R-CNN |