João Paulo Silva Cunha

João Carmona, Tamás Karácsony, João Paulo Silva Cunha

Human motion analysis has seen drastic improvements recently, however, due to the lack of representative datasets, for clinical in-bed scenarios it is still lagging behind. To address this issue, we implemented BlanketGen, a pipeline that augments videos with synthetic blanket occlusions. With this pipeline, we generated an augmented version of the pose estimation dataset 3DPW called BlanketGen-3DPW. We then used this new dataset to fine-tune a Deep Learning model to improve its performance in these scenarios with promising results. Code and further information are available at https://gitlab.inesctec.pt/brain-lab/brain-lab-public/blanket-gen-releases.

João Carmona, Tamás Karácsony, João Paulo Silva Cunha

Clinical in-bed video-based human motion analysis is a very relevant computer vision topic for several relevant biomedical applications. Nevertheless, the main public large datasets (e.g. ImageNet or 3DPW) used for deep learning approaches lack annotated examples for these clinical scenarios. To address this issue, we introduce BlanketSet, an RGB-IR-D action recognition dataset of sequences performed in a hospital bed. This dataset has the potential to help bridge the improvements attained in more general large datasets to these clinical scenarios. Information on how to access the dataset is available at https://rdm.inesctec.pt/dataset/nis-2022-004.

Tamás Karácsony, João Carmona, João Paulo Silva Cunha

Human Pose Estimation (HPE) from monocular RGB images is crucial for clinical in-bed skeleton-based action recognition, however, it poses unique challenges for HPE models due to the frequent presence of blankets occluding the person, while labeled HPE data in this scenario is scarce. To address this we introduce BlanketGen2-Fit3D (BG2-Fit3D), an augmentation of Fit3D dataset that contains 1,217,312 frames with synthetic photo-realistic blankets. To generate it we used BlanketGen2, our new and improved version of our BlanketGen pipeline that simulates synthetic blankets using ground-truth Skinned Multi-Person Linear model (SMPL) meshes and then renders them as transparent images that can be layered on top of the original frames. This dataset was used in combination with the original Fit3D to finetune the ViTPose-B HPE model, to evaluate synthetic blanket augmentation effectiveness. The trained models were further evaluated on a real-world blanket occluded in-bed HPE dataset (SLP dataset). Comparing architectures trained on only Fit3D with the ones trained with our synthetic blanket augmentation the later improved pose estimation performance on BG2-Fit3D, the synthetic blanket occluded dataset significantly to (0.977 Percentage of Correct Keypoints (PCK), 0.149 Normalized Mean Error (NME)) with an absolute 4.4% PCK increase. Furthermore, the test results on SLP demonstrated the utility of synthetic data augmentation by improving performance by an absolute 2.3% PCK, on real-world images with the poses occluded by real blankets. These results show synthetic blanket augmentation has the potential to improve in-bed blanket occluded HPE from RGB images. The dataset as well as the code will be made available to the public.