跳到主要內容

臺灣博碩士論文加值系統

(44.200.140.218) 您好!臺灣時間:2024/07/18 03:50
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

我願授權國圖
: 
twitterline
研究生:康文瑋
研究生(外文):KANG, WEN-WEI
論文名稱:基於遷移式學習及貝茲曲線之棒球轉播智能剪輯系統
論文名稱(外文):Intelligent Pitch Cutting From Professional Baseball Broadcasting using Transfer Learning and Bezier Curve
指導教授:許懷中許懷中引用關係
指導教授(外文):HSU, HWAI-JUNG
口試委員:王益文黃致豪
口試委員(外文):WANG, YI-WENHUANG, JHIH-HAO
口試日期:2020-06-24
學位類別:碩士
校院名稱:逢甲大學
系所名稱:資訊工程學系
學門:工程學門
學類:電資工程學類
論文種類:代替論文:技術報告(應用科技類)
論文出版年:2020
畢業學年度:108
語文別:中文
論文頁數:36
中文關鍵詞:智能剪輯系統電腦視覺遷移式學習貝茲曲線物件辨識機器學習
外文關鍵詞:Intelligent Pitch CuttingComputer VisionTransfer LearningBézier curveObject detectionMachine Learning
相關次數:
  • 被引用被引用:0
  • 點閱點閱:309
  • 評分評分:
  • 下載下載:1
  • 收藏至我的研究室書目清單書目收藏:0
棒球是台灣最興盛的運動產業,每年皆能培養出許多優秀球員投身本國或世界頂級職業賽事,然而近年來,在如奧運、經典賽等頂級國際賽事上,台灣國家隊的表現卻不盡理想,其中一項重要原因便是對於各國在各級職業賽事的職業球員了解不足,傳統的棒球情蒐仍是以人力為主,其中剪輯比賽影片幾乎是所有情蒐分析工作的基礎,然而此項工作耗費大量人力、物力,在資源有限的情況下,經常無法針對比賽對手進行全面的數據分析並訂定出有效的攻守策略。
在本研究中,我們提出一個基於遷移式學習與貝茲曲線的智能剪輯系統,使用少量人工剪輯的影片作為標籤,訓練出可以自動從職業棒球賽事剪輯出投打片段的電腦視覺人工智慧模組。以不同時期和場地進行的賽事進行測試,系統可以將所有投打片段都剪輯出來,且其中平均而言超過九成的片段不需要再經過人工修正,可以大幅度減少剪輯的人力需求。
此外,以上述方法為基礎,我們基於物件辨識模型產生投打畫面的物件特徵,並搭配機器學習模型過濾出投打片段,可以進一步減少情蒐人員審視智能剪輯系統所剪出之投打片段的人力成本。平均而言,我們的系統以 95% 的準確度,節省了 82% 的審視工作,換言之,有這個系統,情蒐人員不再需要人工剪輯影片,僅須檢視回顧 18% 由智能剪輯系統自動檢出的投打片段,就可以完成過去繁重的剪輯工作,可以大幅度提升國家隊、職棒球隊情蒐的效率。
Baseball is the most prosperous sport industry in Taiwan. Every year, lots of excellent talents are grown up and enter the professional leagues around in Taiwan or worldwide. However, in recent years, the national team from Taiwan did not perform well in top international games such as the Olympic or World Baseball Classic. One of the important reasons is our understanding of the players from the other countries is insufficient. Traditionally, baseball intelligence gathering is still human resource intensive, and cutting of baseball films is one of the most essential and time-consuming task in all the intelligence gathering work.
In this thesis, we propose an intelligent cutting system based on transfer learning and Bézier curve. Our approach uses only a few manually edited clips as labels to train an artificial intelligence computer vision module which cut out all the pitching clips from professional baseball game films from different periods or parks. On average, more than 90% of the clips need no further editing, and thus our approach greatly reduces the cutting cost. We also develop another machine learning based method to filter out all the clips which are not perfectly cut for further manual editing. To sum up, our approach saves all the efforts in pitching clip cutting, and reduces 82% review efforts cost with 95% accuracy. In other words, baseball scouts no longer need to edit the clips manually. They need to only review 18% of predicted clips and complete the heavy editing work in the past.

誌謝 i
摘要 ii
Abstract iii
目錄 iv
圖目錄 vi
表目錄 vii
第一章 緒論 1
1.1 研究背景以及動機 1
1.2 研究目的 2
第二章 文獻探討 5
2.1 卷積神經網路 5
2.2 遷移式學習 5
2.3 貝茲曲線 6
2.4 物件辨識 7
2.5 模型可解釋性 7
第三章 系統架構 8
3.1 投打標籤產生方式 8
3.2 自動剪輯系統架構 12
3.3 模型微調 (Fine-tune) 19
3.4 模型的可解釋性 22
3.5 預測結果探討 24
第四章 剪輯結果評估 27
第五章 結論以及未來工作 31
參考文獻 33

[1]謝仕淵; 謝佳芬. 台灣棒球一百年. 果實出版社, ISBN: 9867796101. 2003/08.
[2]曾文誠; 盂峻瑋. 台灣棒球王. 我識出版社, ISBN: 9572944983. 2004/06.
[3]何信弘. 中華台北棒球代表隊情蒐工作之質性研究 (2001~ 2009), 碩士論文, 國立台北教育大學, 臺北市. 2008.
[4]張厥煒; 林華韋. 雅典奧運棒球情蒐資訊系統建構與運作. 大專體育, 2005, 76: 25-30.
[5]張厥煒; 楊清瓏. 2008 年北京奧運中華隊棒球情蒐工作之回顧與展望.
中華體育季刊, 2009, 23.4: 141-149
[6]Chiu, Chih-Yi, et al. "Detecting pitching frames in baseball game video using markov random walk." 2010 IEEE International Conference on Image Processing. IEEE, 2010.
[7]Zhang, Dongqing, and Shih-Fu Chang. "Event detection in baseball video using superimposed caption recognition." Proceedings of the tenth ACM international conference on Multimedia. 2002.
[8]Redmon, Joseph, et al. "You only look once: Unified, real-time object detection." Proceedings of the IEEE conference on computer vision and pattern recognition. 2016.
[9]Redmon, Joseph, and Ali Farhadi. "Yolov3: An incremental improvement." arXiv preprint arXiv:1804.02767 (2018).
[10]Bochkovskiy, Alexey, Chien-Yao Wang, and Hong-Yuan Mark Liao. "YOLOv4: Optimal Speed and Accuracy of Object Detection." arXiv preprint arXiv:2004.10934 (2020).
[11]Russakovsky, Olga, et al. "Imagenet large scale visual recognition challenge." International journal of computer vision 115.3 (2015): 211-252.
[12]Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton. "Imagenet classification with deep convolutional neural networks." Advances in neural information processing systems. 2012.
[13]Howard, Andrew G. "Some improvements on deep convolutional neural network based image classification." arXiv preprint arXiv:1312.5402 (2013).
[14]Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton. "Imagenet classification with deep convolutional neural networks." Advances in neural information processing systems.
[15]M. Lin, Q. Chen, and S. Yan. "Network in network." arXiv:1312.4400,2013.
[16]Srivastava, Nitish, et al. "Dropout: a simple way to prevent neural networks from overfitting." Journal of Machine Learning Research 15.1 (2014): 1929-1958.
[17]Hinton, Geoffrey E., et al. "Improving neural networks by preventing co- adaptation of feature detectors." arXiv preprint arXiv:1207.0580 (2012).
[18]Glorot, Xavier, Antoine Bordes, and Yoshua Bengio. "Deep Sparse Rectifier Neural Networks." Aistats. Vol. 15. No. 106. 2011.
[19]Pan, Sinno Jialin, and Qiang Yang. "A survey on transfer learning." IEEE Transactions on knowledge and data engineering 22.10 (2009): 1345-1359.
[20]Simonyan, Karen, and Andrew Zisserman. "Very deep convolutional networks for large-scale image recognition." arXiv preprint arXiv:1409.1556 (2014).
[21]Szegedy, Christian, et al. "Going deeper with convolutions." Proceedings of the IEEE conference on computer vision and pattern recognition. 2015.
[22]Szegedy, Christian, et al. "Inception-v4, inception-resnet and the impact of residual connections on learning." Thirty-first AAAI conference on artificial intelligence. 2017.
[23]Yosinski, Jason, et al. "How transferable are features in deep neural networks?." Advances in neural information processing systems. 2014.
[24]Liu, Wei, et al. "Ssd: Single shot multibox detector." European conference on computer vision. Springer, Cham, 2016.
[25]Girshick, Ross, et al. "Rich feature hierarchies for accurate object detection and semantic segmentation." Proceedings of the IEEE conference on computer vision and pattern recognition. 2014.
[26]Girshick, Ross. "Fast r-cnn." Proceedings of the IEEE international conference on computer vision. 2015.
[27]Ren, Shaoqing, et al. "Faster r-cnn: Towards real-time object detection with region proposal networks." Advances in neural information processing systems. 2015.
[28]Zhou, Kaiyang, et al. "Omni-scale feature learning for person re-identification." Proceedings of the IEEE International Conference on Computer Vision. 2019.
[29]Zeiler, Matthew D., and Rob Fergus. "Visualizing and understanding convolutional networks." European conference on computer vision. Springer, Cham, 2014.
[30]Zhou, Bolei, et al. "Learning deep features for discriminative localization." Proceedings of the IEEE conference on computer vision and pattern recognition. 2016.
[31]Selvaraju, Ramprasaath R., et al. "Grad-cam: Visual explanations from deep networks via gradient-based localization." Proceedings of the IEEE international conference on computer vision. 2017.
[32]中華職棒CPBL TV. (https://www.cpbltv.com/)
[33]中華職棒大聯盟全球資訊網. The Official Site of CPBL. (http://www.cpbl.com.tw/)
[34]Bradski, G. (2000). The OpenCV Library. Dr. Dobb's Journal of Software Tools.

電子全文 電子全文(本篇電子全文限研究生所屬學校校內系統及IP範圍內開放)
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
無相關期刊