近年來，運動科技逐漸發展，選手訓練過程主要搭配攝影設備以及穿戴式裝置，而游泳池也需要搭配前述專業設備建置，用以提供相關分析數據給游泳教練，以科學化的方式來得到更好的訓練成果。目前市面上有攝影以及穿戴式裝置的整合動作捕捉系統，主要用以記錄跳水以及游泳過程中選手的位置資訊，然而，此類系統的價格不斐，設置上也需要花費大量的時間。本研究提供了跳水以及游泳兩套分析系統，在使用上省去了上述繁瑣操作以及所需金錢成本，只要使用單一攝影機所拍攝之2D RGB影片即可估計影片中選手的人體骨架位置資訊，其功能包含了教練所需要的數據計算，例如髖關節角度、速度、跳水距離、滯空時間以及划頻划距等，也可以提供教練觀看選手的運動軌跡。兩套分析系統皆使用了人體骨架估計演算法Openpose之改良版，我們藉由減少PAF階段並且加入Densenet架構的方式，來縮小其尺寸並維持骨架關節位置的預測精準度，以節省模型訓練和人體骨架估計所需的時間，訓練資料則是根據Youtube以及教練提供的影片來製作。接著，根據模型給出的骨架估計結果，針對數種錯誤情形(例如、部分關節點消失問題)提出修正演算法，來得到比較準確的骨架關節點位置資訊，以利於後續量化數據的產生。

Recently, sports technology gradually develops and changes the training process in swimming. With the assistance of professional photographic equipment and wearable devices, the design of the swimming pool is gradually moving towards intelligence. The integrated system of wearable devices and photographic equipment is available and used to make quantitative statics during training process like in-water pose, velocity and angle of swimmer. Although this kind of system is useful for swimmer’s performance analysis, the cost of time and money is expensive. In this study, we provide two analysis systems for diving and swimming respectively, which eliminates the cumbersome setup and cost mentioned above. We can only use 2D RGB video recorded by single camera to estimate the positions of skeleton joints. These two systems include the data required by the coach, such as hip angle, velocity, diving distance, airtime, frequency, and distance of swimming stroke. Also, both systems provide the trajectory of athlete. Both analysis systems use a modified version of Openpose. By adding Densenet architecture and decreasing the number of PAF stages, the size of model can be reduced. We can save the model training and testing time. The precision of skeleton joint position estimation is maintained. The training data is based on swimming coach and Youtube. Then, according to the estimation of skeleton given by the model, we provide correction algorithm to deal with several error cases(e.g. partial joint missing problem) to obtain more accurate skeleton joint location information for the subsequent quantitative data computation.

[1] S.E. Wei, V. Ramakrishna, T. Kanade, and Y. Sheikh, “Convolutional Pose Machines,” In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 4724-2732, 2016.
[2] Z. Cao, G.H. Martinez, T. Simon, S.E. Wei, and Y.A. Sheikh, “OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 43, no. 1, pp. 172-186, 2021.
[3] G.H. Martinez, Y. Raaj, H. Idrees, D. Xiang, H. Joo, T. Simon, and Y. Sheikh, “Single-Network Whole-Body Pose Estimation,” In Proceedings of the IEEE International Conference on Computer Vision, pp. 6982-6991, 2019.
[4] Z. Zheng, T. Yu, Y. Wei, Q. Dai, and Y. Liu, “DeepHuman: 3D Human Reconstruction From a Single Image,” In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 7739-7749 , 2019.
[5] D. Mehta, O. Sotnychenko, F. Mueller, W. Xu, M. Elgharib, P. Fua, H.P. Seidel, H. Rhodin, G. Pons-Moll, and C. Theobalt, “XNect: real-time multi-person 3D motion capture with a single RGB camera,” ACM Transactions on Graphics, vol. 39, no. 4, pp. 82:1-82:17, 2020.
[6] V. Bazarevsky, I. Grishchenko, K. Raveendran, T. Zhu, F. Zhang, and M. Grundmann, “BlazePose: On-device Real-time Body Pose tracking,” ArXiv preprint arXiv:2006.10204, 2020.
[7] Y. Raaj, H. Idrees, G. Hidalgo, and Y. Sheikh, “Efficient Online Multi-Person 2D Pose Tracking with Recurrent Spatio-Temporal Affinity Fields,” In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 4620-4628, 2019.
[8] B. Xiao, H. Wu, Y. Wei, “Simple Baselines for Human Pose Estimation and Tracking,” In Proceedings of 15th European Conference on Computer Vision, pp. 466-481, 2018
[9] G. Ning, J. Pei, H. Huang, “LightTrack: A Generic Framework for Online Top-Down Human Pose Tracking,” In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1034-1035,2020.
[10] D. Osokin, “Real-time 2D Multi-Person Pose Estimation on CPU: Lightweight OpenPose,” In Proceedings of 8th International Conference on Pattern Recognition Applications and Methods, pp 744-748, 2019
[11] L. Bertinetto, J. Valmadre, J.F. Henriques, A. Vedaldi, and P.H.S. Torr, “Fully-Convolutional Siamese Networks for Object Tracking,” In Proceedings of 14th European Conference on Computer Vision, pp. 850-865, 2016.
[12] B. Li, J. Yan, W. Wu, Z. Zhu, and X. Hu, “High Performance Visual Tracking With Siamese Region Proposal Network,” In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 8971-8980, 2018.
[13] B. Li, W. Wu, Q. Wang, F. Zhang, J. Xing, and J. Yan, “SiamRPN++: Evolution of Siamese Visual Tracking With Very Deep Networks,” In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 4277-4286, 2019.
[14] Q. Wang, L. Zhang, L. Bertinetto, W. Hu, and P.H.S. Torr, “Fast Online Object Tracking and Segmentation: A Unifying Approach,” In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1328-1338, 2019.
[15] K. Sun, B. Xiao, D. Liu, and J.D. Wang, “Deep High-Resolution Representation Learning for Human Pose Estimation,” In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 5693-5703, 2019.
[16] G. Papandreou, T. Zhu, N. Kanazawa, A. Toshev, J. Tompson, C. Bregler, and K. Murphy, “Towards Accurate Multi-Person Pose Estimation in the Wild,” In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 4903-4911, 2017.
[17] S. Kreiss, L. Bertoni, and A. Alahi, “PifPaf: Composite Fields for Human Pose Estimation,” In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 11977-11986, 2019.
[18] K. Toyoda, M. Kono, and J. Rekimoto, “Post-Data Augmentation to Improve Deep Pose Estimation of Extreme and Wild Motions,” In IEEE Conference on Virtual Reality and 3D User Interfaces (VR), pp. 1570-1574, 2019.
[19] A. Viswakumar, V. Rajagopalan, T. Ray and C. Parimi, “Human Gait Analysis Using OpenPose,” In Proceedings of Fifth International Conference on Image Information Processing (ICIIP), pp. 310-314, 2019.
[20] T.Y. Lin, M. Maire, S. Belongie, L. Bourdev, R. Girshick, J. Hays, P. Perona, D. Ramanan, C.L. Zitnick, and P. Dolla ́r, “Microsoft COCO: common objects in context,” In Proceedings of 13th European Conference on Computer Vision, pp. 740-775, 2014.
[21] G. Huang, Z. Liu, K.Q. Weinberger, and L. Maaten, “Densely connected convolutional networks,” In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 4700-4708, 2017.
[22] Y. Zhang, Y. Tian, Y. Kong, B. Zhong, Y. Fu, “Residual Dense Network for Image Super-Resolution, In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2472-2481, 2018.
[23] M. Unser, "Splines: a perfect fit for signal and image processing," In IEEE Signal Processing Magazine, vol. 16, no. 6, pp. 22-38, 1999.
[24] A.C.S. Guimaraes, and J.G. Hay, “A Mechanical Analysis of the Grab Starting Technique in Swimming,” Journal of Applied Biomechanics, vol. 1, no. 1, pp. 25-35, 1985.
[25] J. Vantorre, C. Didier, and L. Seifert,“Biomechanical Analysis of the Swim-Start: A Review,” Journal of sports science & medicine. vol. 13, no.2, 223-231, 2014.