Resnet-BiLSTM Network for Multi-Label Weather Image Classification

Weather phenomena usually contained multiple basic weather elements. To tackle the problem of multi-label weather classification, an algorithm, mainly consist of feature extraction module based on Resnet34, classification module based on BiLSTM, is proposed in this paper. To demonstrate the effectiveness of the proposed algorithm, some experiments are performed on a multi-label weather database, including five classes. The proposed algorithm scores mAP 89.6, CP 85.4, CR 76.5, the experimental results show that the proposed algorithm performances pretty good on multi-label weather classification tasks.