In this paper, a face retrieval method based on deep neural networks with relevance feedback has been presented to solve the problem of large-scale human face retrieval. Firstly, convolutional neural networks has been used for feature extracting, then, traditional search methods will be used in face retrieval. A feedback model is added after the retrieval stage. According to the users' feedback, result samples are divided into positives and negatives, which are be used for training the feedback model. Experiments show that this method can significantly improve the accuracy of face retrieval.