本研究使用Faster R-CNN物件偵測演算法，在TensorFlow平臺上構建深度學習模型，用以識別中共戰鬥機；在這項研究中，我們試圖探究Faster R-CNN物件偵測演算法運用數種神經網路層級對於中共戰鬥機辨識的效能。研究結果發現，使用ResNet-101神經網路層的Faster R-CNN物件偵測演算法優於使用Inception V2神經網路層的Faster R-CNN物件偵測演算法。此外，Faster R-CNN物件偵測演算法使用ResNet-101神經網路層的中共戰鬥機辨識效能優於使用ResNet-50神經網路層；最後，本研究發現，增加Faster R-CNN物件偵測演算法的訓練次數可以提高單一型別中共戰鬥機的識別率；然而，這可能不會提高多型別中共戰鬥機的識別率。

This study used the Faster R-CNN object detection algorithms to build a deep learning model on a TensorFlow platform to identify P.R.O.C. fighters. We tried to explore the PROC fighter identification performance of the Faster R-CNN object detection algorithm with several neural network layers in this study. Regarding the P.R.O.C. fighter identification, the study results found that the Faster R-CNN object detection algorithm with the ResNet-101 neural network layer outperforms the Faster R-CNN object detection algorithm with the Inception V2 neural network layer. Also, the Faster R-CNN object detection algorithm with the ResNet-101 neural network layer receives better performance than the Faster R-CNN object detection algorithm with the ResNet-50 neural network layer. Finally, this study found that increasing the training times of the Faster R-CNN object detection algorithm can improve the identification rate of a single type PROC fighter; however, it might not improve the identification rate of multiple type PROC fighters.