智慧眼鏡是一種穿戴式技術，正在改變我們與世界互動的方式。本研究旨在利用深度學習技術，特別是基於YOLO（You Only Look Once）的物體檢測方法，實現在智慧眼鏡上的實時物體檢測和識別。這項研究結合了計算機視覺、機器學習和穿戴式技術領域的前沿知識，在拓展智慧眼鏡的應用範疇，提高其實用性和互動性。

    我們首先簡要介紹了智慧眼鏡和YOLO物體檢測的背景和原理。隨後，我們描述了實驗設計和數據收集過程，包括標註和預處理訓練數據的步驟。我們利用YOLOv5模型進行實驗，該模型在物體檢測任務中表現優異，具有高速度和高效率的特點。實驗結果顯示，智慧眼鏡檢測模型能夠實時檢測和識別周圍環境中的多個物體。

    這項研究的成功應用為智慧眼鏡的實用性增添了新的維度，可能在輔助導航、視覺搜索、工業應用和安全監控等多個領域發揮關鍵作用。然而，我們也提出了一些挑戰和未來工作的建議，包括性能優化、隱私保護和用戶體驗的改進。深度學習在穿戴式技術中的應用將繼續受到關注，並為智慧眼鏡等設備的發展帶來更多的創新和可能性。

Smart glasses are wearable technology that is changing the way we interact with the world. This research aims to use deep learning technology, especially the object detection method based on YOLO (You Only Look Once), to achieve real-time object detection on smart glasses. This research combines the forefront of computer vision, machine learning and wearable technology. Knowledge, in expanding the application field of intelligent vision, improving practicability and interactivity.

 We first briefly introduce the background and principles of smart glasses and YOLO object detection. We then describe the experimental design and data collection process, including steps for supplementing and training data. We use the YOLOv5 model to conduct experiments, which has excellent performance in target detection tasks and has the characteristics of high speed and high efficiency. Experimental results show that the smart glasses detection model can detect and recognize multiple objects in the surrounding environment in real time.

 The successful application of recent research adds a new dimension to the practicality of eyewear intelligence, which may play a key role in areas as diverse as assisted navigation, visual search, industrial applications, and security monitoring. However, we also propose some challenges and suggestions for future work, including performance optimization, privacy protection, and user experience improvement. The application of deep learning in wearable technology will continue to attract attention and bring more innovations and possibilities to the development of smart glasses and other devices.

目錄
誌謝	I
摘要	II
ABSTRACT	III
目錄	I
圖目錄	III
表目錄	IV
第一章	緒論	1
1.1　研究背景	1
1.2　研究動機與目的	2
1.3　研究流程	3
1.4　研究架構	3
第二章	文獻探討	5
2.1　智慧裝置	6
2.2　影像辨識與模型評估	7
2.3　資料蒐集	9
第三章	研究方法	12
3.1　資料前處理	12
3.2　研究內容	13
第四章	研究結果	18
4.1　智慧眼鏡物件偵測實驗	19
4.2　實驗設計	20
4.3　實驗評估方式	21
4.4　實驗結果	21
第五章	結論與建議	23
5.1　研究結論	23
5.2　建議與未來發展	23
參考文獻	24

圖目錄
圖 1　研究流程	4
圖 2　物件辨識架構表	7
圖 3　LabelImg 工具示意圖	12
圖 4　J7EF-Gaze智慧眼鏡	13

表目錄
表 1　國外相關論文/期刊整理	10
表 2　國內論文/期刊整理	11
表 3　各模型訓練條件表	15
表 4　YOLOv4 tiny實驗結果	17
表 5　Scaled-YOLOv4實驗結果	17
表 6　YOLOR實驗結果	17
表 7　YOLOv5實驗結果	17
表 8　綜合比較表	18
表 9　智慧眼鏡規格[12]	19
表 10　智慧眼鏡物件偵測實驗範本	20
表 11　智慧眼鏡物件偵測實驗結果	21
表 12　影像預測結果	22
 

[1] A. Bochkovskiy, C. Y. Wang, and H. Y. Mark Liao,“YOLOv4：Optimal speed and accuracy of object detection,” in arXiv, arXiv:2004.10934, 2020.
[2] Ross Girshick, “Fast R-CNN," arXiv:1504.08083v2, 2015 
[3] Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi,“You Only Look Once:Unified, Real-Time Object Detection,”  in IEEE Conference on Computer Vision and Pattern Recognition, 2016.
[4] Joseph Redmon and Ali Farhadi, “YOLOv3: An Incremental Improvement,” arXiv:1804.02767v1, 2028
[5] Joseph Redmon and Ali Farhadi, “YOLO9000: Better, Faster, Stronger,” arXiv:1612.08242v1, 2016
[6] Wei Liu, Dragomir Anguelov, Dumitru Erhan, Christian Szegedy,Scott Reed, Cheng-Yang Fu, and Alexander C. Berg, “SSD: Single Shot MultiBox Detector,” in Computer Vision and Pattern Recognition, 2016.
[7] LabelImg Github. 
   Available: https://github.com/HumanSignal/labelImg Sep 23, 2022
[8] roboflow官方網站.
   Available: https://models.roboflow.com/object-detection/yolov4-tiny-darknet
[9] roboflow官方網站.
   Available: https://models.roboflow.com/object-detection/scaled-yolov4
[10] roboflow官方網站.
   Available: https://models.roboflow.com/object-detection/yolor
[11] roboflow官方網站.
   Available: https://models.roboflow.com/object-detection/yolov5 
[12] Jorjin. Available: https://www.jorjin.com/