CBL: A Clothing Brand Logo Dataset and a New Method for Clothing Brand Recognition
Kuan-Hsien Liu, Tsung-Jung Liu, Fei Wang
Abstract
In this work, we presented a novel clothing brand logo prediction method which is rooted on a dense-block based deep convolutional neural network for brand logo detection and recognition. To learn convolutional neural networks deeper and more accurately, we adopted dense blocks into deep convolutional networks to make connections between layers shorter. In our work, we propose several dense-block structure designs to improve detection and recognition accuracy on clothing brand logos. We also built a new large-scale clothing brand logo (CBL) dataset with the brand attribute and logo information to facilitate this task. To reduce complexity for the proposed framework, two pixel search steps for the bounding movement are implemented in the training procedure. In the experiment, we demonstrate our search reduced model can outperform some state-of-the-art methods and achieve very good results.