Clothing Brand Prediction via Logo Recognition

To deeper and more accurately to learn convolutional neural networks, dense blocks can be incorporated into convolutional networks to shorten connections between layers. In this paper, we proposed a new clothing brand prediction method based on a dense-block deep convolutional neural network for logo detection and recognition. Several dense blocks are designed to improve prediction accuracy on clothing brand logo. We also constructed a new clothing dataset with brand and logo information to facilitate this task. In the experiment, we show our method can achieve better performance than some state-of-the-art methods.