Aquatic Ecology ( IF 1.8 ) Pub Date : 2023-12-12 , DOI: 10.1007/s10452-023-10078-y A. Carneiro , L. S. Nascimento , M. A. Noernberg , C. S. Hara , A. T. R. Pozo
The Portuguese man-of-war is responsible for the most common and severe stings worldwide. Jellyfish monitoring is essential to manage stings, and social media is a valuable data source for obtaining observations of this species. This study reports on using Convolutional Neural Networks for Portuguese man-of-war image classification extracted from social media posts. We created a suitable dataset and trained three different neural networks: VGG-16, ResNet50, and InceptionV3, with and without a pre-trained step with the ImageNet dataset. The pre-trained ResNet50 network presented the best results, obtaining 94% accuracy and 95% precision, recall, and F1 score. We conclude that Convolutional Neural Networks can be very effective for recognizing Portuguese man-of-war images from social media, helping in obtaining data about its occurrence and distribution.
中文翻译:
用于水母监测的社交媒体图像分类
葡萄牙战舰是世界上最常见、最严重的蜇伤事件的罪魁祸首。水母监测对于管理水母蜇伤至关重要,而社交媒体是获取该物种观察结果的宝贵数据源。这项研究报告了如何使用卷积神经网络对从社交媒体帖子中提取的葡萄牙军舰图像进行分类。我们创建了一个合适的数据集,并训练了三种不同的神经网络:VGG-16、ResNet50 和 InceptionV3,无论是否使用 ImageNet 数据集进行预训练步骤。预训练的 ResNet50 网络呈现出最好的结果,获得了 94% 的准确率和 95% 的精确度、召回率和 F1 分数。我们的结论是,卷积神经网络可以非常有效地识别社交媒体上的葡萄牙军舰图像,有助于获取有关其发生和分布的数据。