Abstract: Detection of long-rang underwater acoustic targets is a challenging task due to low signal-to-noise ratio (SNR) of the target radiation noise. In order to improve the detection accuracy of long-range underwater acoustic targets, a method based on the densely connected neural network and the attention mechanism is proposed. The Mel-frequency cepstral coefficients are extracted as features, a self-attentive module is added to the head of the densely connected neural network to capture key information, and the detection results are obtained after several dense blocks. The experiments are conducted on the real-world dataset, and the performance variations of the model are separately analyzed whether the attention mechanism is added and when the input characteristics and the receiver depth are different. In testing the model with data from the next few days, the detection accuracy is 93.3% for the detection range less than 10 km, and 90.34% for 20 km. The experimental results show that this model achieves the underwater acoustic target detection at SNR not less than −6 dB and still has the detection capability after adding more low SNR samples, thus it performs better than the other models. In addition, the detection performance can be further improved when the training set contains data in various SNR conditions.