【摘要】 本文提出了一种基于Mask-RCNN和数据集DeepFashion2的服装识别与分割的方法。基于Mask-RCNN的服装识别与分割是基于卷积神经网络的思想,在深度学习框架下通过多线程迭代训练,在ResNet网络中得到目标特征后,再通过RPN和RoI Align将特征输入不同的全连接分支,最后得到具有优化权重的目标检测模型。在不同场景的服装图像中,该模型可以更快更准确的识别出服装并将其分割。更多还原

【Abstract】 A method for clothing recognition and segmentation based on Mask-RCNN and data set DeepFashion2 was proposed.Mask-RCNN-based clothing recognition and segmentation was based on the idea of convolutional neural networks.Through the multithreaded iterative training in the deep learning framework,after obtaining the target features in the ResNet network,the features were input differently through RPN and RoI Align.The branches were connected.Finally the target detection model with optimized weights was obtained.In clothing image of different scenes,the model could identify and segment the garment more quickly and accurately.更多还原