This repository includes scripts for preprocessing videos to generate pickle files and training.
Download the wheelchair keypoints dataset for human pose estimation through this link.
Download samples from the SAFER Activities Dataset through this link.
Download the pickle files for training models using keypoint information through this link.
Download example label visualizations on videos through this link.
The repository is organized into three main folders:
-
preprocessing
: Contains code for preprocessing videos to generate pickle files with keypoints. We used the ViTPose-h-multi and Yolov8x models for our preprocessing. -
keypoints_train
: Contains code to train our Simple-1D-CNN. This folder includes all necessary scripts to start the training process and evaluate the model's performance on our dataset. -
inference
: Contains code to run inference, as well as run evalation on the SAFER-Activities test set. Additionally includes evaluation code to run inference on other realistic fall datasets. -
docker
: Includes aDockerfile
andrun_container.sh
script.
This dataset is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0). This means you can share and adapt the material for any non-commercial purpose as long as you provide appropriate credit, include a link to the license, and indicate if changes were made. You must also share any adaptations under the same license.