本論文設計並實作一套人臉辨識系統，結合嵌入式裝置、跨平台手機應用程式與深度學習模型，應用於協助提醒與分辨眼前的人。目前人臉辨識的應用多用於公司、商場等大型應用領域，尚未有應用於個人方面的產品，本論文所設計的系統將人臉辨識系統個人化，應用於和以往不同的場景。本系統主要功能包含自動從已記錄下的人中比對並告知使用者目前眼前最可能是記錄中哪一個人，以及記錄新人的照片，讓使用者之後再對其做註記。
本論文使用嵌入式系統連接鏡頭做為影像輸入裝置，並開發跨平台的手機應用程式。在人臉辨識流程中，使用ML Kit進行人臉偵測，以及使用MobileNetV2以及MobileFaceNet兩個深度學習模型輸出人臉特徵，並比較了這兩個模型在手機上執行的表現，得出MobileFaceNet在手機上執行人臉辨識有較好的表現。在無線連線上採用了藍牙低功耗的方式，同時為了解決藍牙低功耗傳輸資料大小上的限制，也設計了相關的大型資料傳送演算法解決問題，並比較了傳輸不同大小與品質的圖片對整個系統的運作流程的影響。最後，本系統將系統的各流程串連，實做出自動記錄陌生人、使用者標記頁面，已知的人偵測與提醒等功能，同時也能達到手機獨立運作，與保留了嵌入式裝置獨立運作的擴充功能。在實際運作時，本系統可以達到94% 的人臉辨識正確率，以及1 FPS的辨識速度。

This thesis designs and implements a face recognition system, which combines an embedded system, cross-platform mobile applications and deep learning models to help a user to check if the user knows or has met the people in front of the user. At present, there is no applications of face-recognition similar to the proposed approach to help individuals.
The main functions of this system include automatically comparing the captured picture with the recorded people and notifying the user which person is most likely to be, automatically inserting the not-recorded persons encountered into the records, and user annotation.
This thesis uses an embedded system to connect a camera as the image input device, and uses Flutter to develop a cross-platform mobile application. In the face recognition process, ML Kit is used for face detection, and MobileFaceNet is adopted after comparison with MobileNetV2. Wireless connection is BLE. To solve the limitation of the data size of BLE transmission, a related large-scale data transmission algorithm is also implemented to solve the problem. Finally, this face recognition system achieves 94% recognition accuracy and 1 FPS recognition speed.

[1] Alexey Egorov, “Algorithm for optimization of Viola–Jones object detection framework parameters” Journal of Physics Conference Series, vol. 945, no. 1, p. 012032, DOI:10.1088/1742-6596/945/1/012032, Jan. 2018.
[2] R. Girshick, "Fast R-CNN", IEEE International Conference on Computer Vision, pp. 1440-1448, Apr. 2015.
[3] Joseph Redmon, Santosh Divvala, Ross Girshick and Ali Farhadi, “You Only Look Once: Unified, Real-Time Object Detection”, IEEE Conference on Computer Vision and Pattern Recognition, pp. 779-788, Dec. 2016.
[4] L. Lai, N. Suda and V. Chandra, “CMSIS-NN: Efficient Neural Network Kernels for Arm Cortex-M CPUs”, arXiv preprint arXiv:1801.06601, 2018.
[5] Y. Taigman, M. Yang, M. Ranzato and L. Wolf, “DeepFace: Closing the Gap to Human-Level Performance in Face Verification” IEEE Conference on Computer Vision and Pattern Recognition, pp. 1701-1708, 2014.
[6] Andrew G. Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco Andreetto and Hartwig Adam, “MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications”, arXiv preprint arXiv:1704.04861, 2017.
[7] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov and L. C. Chen, “MobileNetV2: Inverted Residuals and Linear Bottlenecks”, IEEE Conference on Computer Vision and Pattern Recognition, pp. 4510-4520, 2018.
[8] S. Chen, Y. Liu, X. Gao and Z. Han, “Mobilefacenets: Efficient cnns for accurate real-time face verification on mobile devices”, Chinese Conference on Biometric Recognition, pp. 428-438, 2018.
[9] F. Schroff, D. Kalenichenko and J. Philbin, “FaceNet: A unified embedding for face recognition and clustering”, IEEE Conference on Computer Vision and Pattern Recognition, pp. 815-823, 2015.
[10] J. Deng, J. Guo, N. Xue and S. Zafeiriou, “ArcFace: Additive Angular Margin Loss for Deep Face Recognition”, IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 4685-4694, 2019.
[11] Flutter Developer Tools, [Online]. Available: https://flutter.dev, last retrieve: 18 August 2022.
[12] Google ML Lit, [Online]. Available: https://developers.google.com/ml-kit, last retrieve: 18 August 2022.
[13] How is Flutter different for app development, [Online]. Available: https://www.youtube.com/watch?v=l-YO9CmaSUM, last retrieve: 18 August 2022.
[14] Flutter text to speech package, [Online]. Available: https://pub.dev/packages/flutter_tts, last retrieve: 18 August 2022.
[15] 胡宇謙, 設計與實作便於攜帶的即時人臉辨識系統, 碩士論文，成功大學工程科學系, 2022
[16] 邱子懿, 使用可應用於行動裝置之臉部辨識神經網路進行年齡預測與分類之實作, 碩士論文，成功大學工程科學系, 2022.
[17] Dong Yi, Zhen Lei, Shengcai Liao and Stan Z. Li, “Learning Face Representation from Scratch”, arXiv preprint arXiv:1411.7923, 2014.
[18] NodeJS camera package, [Online]. Available: https://www.npmjs.com/package/v4l2camera, last retrieve: 18 August 2022.
[19] NodeJS image processing library, [Online]. Available: https://www.npmjs.com/package/jimp, last retrieve: 18 August 2022.
[20] NodeJS Bluetooth socket package, [Online]. Available: https://www.npmjs.com/package/bluetooth-hci-socket, last retrieve: 18 August 2022.
[21] Raspberry Pi Zero W, [Online]. Available: https://www.raspberrypi.com/products/raspberry-pi-zero-w, last retrieve: 18 August 2022.