This repository is the official implementation of Detecting Arbitrary Keypoints on Limbs and Skis with Sparse Partly Correct Segmentation Masks.
The paper is accepted for CV4WS'23 at WACV'23.
You can download the dataset here.
After downloading, you need to download the videos from YouTube and extract the annotated frames.
The annotated frames are expected to be located in a folder has subfolders for each video (named by its index)
and each subfolder contains the frames with the following naming convention: <video_id>_(<frame_num>).jpg,
whereby the frame num is five-digit, e.g.<frames_path>/0/0_(01944).jpg.
Install all requirements in requirements.txt.
Before running, you need to adjust the paths in paths.py. You need to set the path to the extracted frames,
the path to the downloaded folder, the weights path and a file where the inference results will be dumped.
The pretrained weights are located in model/pretrained/model_seg_proj.pth.tar.
If you want to use another GPU than 0, adjust it in the header of evaluation/eval_yt_skijump.py. Then, run this file:
python -m evaluation.eval_yt_skijump.pyIn case this work is useful for your research, please consider citing:
@InProceedings{ludwig2023arbitrary_kps,
title = {Detecting Arbitrary Keypoints on Limbs and Skis with Sparse Partly Correct Segmentation Masks},
author = {Ludwig, Katja and Kienzle, Daniel and Lorenz, Julian and Lienhart, Rainer},
booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
month = {January},
year = {2023},
}
Our code is heavily influenced by the following projects:
We are grateful to the corresponding authors for releasing their code. We use some of their code in our project.