Abstract | ||
---|---|---|
In this paper, we propose a novel loss function named dual-guided loss (DGL) for ground-based cloud classification in weather station networks. The proposed DGL could integrate the knowledge of different convolutional neural networks (CNNs) in the process of optimization, which improves the discriminative ability of ground-based cloud feature representations. To this end, we add a modulation term into the DGL, which assigns large weights to the hard-classified ground-based cloud samples. As a result, the deep network is forced to pay more attention to these hard-classified samples, and therefore, the performance of the deep network gets improved. We demonstrate the effectiveness of the proposed DGL with the extensive experiments on two ground-based cloud datasets, and the experimental results of the DGL outperform the state-of-the-art methods. |
Year | DOI | Venue |
---|---|---|
2019 | 10.1109/ACCESS.2019.2916905 | IEEE ACCESS |
Keywords | Field | DocType |
Dual guided loss,ground-based cloud classification,weather station networks | Computer science,Weather station,Computer network,Real-time computing,Cloud computing | Journal |
Volume | ISSN | Citations |
7 | 2169-3536 | 0 |
PageRank | References | Authors |
0.34 | 0 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Mei Li | 1 | 2 | 11.53 |
Shuang Liu | 2 | 36 | 22.95 |
Zhong Zhang | 3 | 141 | 32.42 |