中文版:Mona：Mona 數字記憶網路
我們提出了用於從影象中識別人臉的神經網路 (NN)。當對人類面部表情進行訓練時，該網路表現出自然的和判別性的判別特徵，這些特徵是透過將兩個深度神經網絡合二為一來提取的。神經網路專注於面部屬性的識別，而深度神經網路則使用面部照片的特徵表示進行訓練。我們透過學習具有透過訓練和測試學習的相同特徵表示的模型表示，開發了一種新的人臉識別方法。我們表明訓練中的訓練樣本不足以處理多檢視人臉識別。我們針對這個問題提出了兩種變體，並提出了一種使用深度卷積神經網路 (CNN) 進行人臉識別的替代方法。CNN 模型適用於人臉識別任務，從而避免了對人臉屬性進行建模的需要。最後，我們表明 CNN 透過利用訓練樣本中的判別特徵來提高效能。我們討論了人臉識別可用於改進影象檢索的應用。
英文原版:Mona: The Mona Digital Memory Network
We present the Neural Network (NN) for face recognition from images. When trained on human facial expressions, the Network exhibits natural and discriminative discriminative feature features which are extracted by combining two deep neural networks into one. NNs focus on the recognition of facial attributes, whereas deep neural networks are trained using the feature representation of face photographs. We develop a novel approach to face recognition by learning the model representations with the same feature representations learned by training and test. We show that the training samples in training are not large enough to handle multi-view face recognition. We present two variations on this problem and propose an alternative approach for face recognition using the deep Convolutional Neural Network (CNN). The CNN model is adapted to the task of face recognition, thus avoiding the need to model facial attributes. Finally we show that CNN improves perf 這些內容均是人工智慧自動生成，不是真實存在。
