NorFisk Dataset

Long-term autonomous monitoring of wild fish populations surrounding fish farms can contribute to a better understanding of interactions between wild and farmed fish, which can have wide-ranging implications for disease transmission, stress in farmed fish, wild fish behavior and nutritional status, etc. The ability to monitor the presence of wild fish and its variability with time and space will improve our understanding of the dynamics of such interactions and the implications that follow. Many efforts are underway to recognize fish species using artificial intelligence. However there are not many image datasets publicly available to train these neural networks, and even fewer that include species that are relevant for the aquaculture sector. Here we present a public dataset of annotated images for fish species recognition with deep learning. The dataset contains 9487 annotated images of farmed salmonids and 3027 annotated images of saithe and it is expected to grow in the near future. This dataset was the result of processing nearly 50 hours of video footage taken inside and outside cages from several fish farms in Norway. The footage was processed with a semi-automatic system to create large image datasets of fish under water. The system combines techniques of image processing with deep neural networks in an iterative process to extract, label, and annotate images from video sources. The details of the system are described in a journal paper that is currently under review. This information will be updated when the paper is published.

对养鱼场周边野生鱼类种群开展长期自主监测，可助力我们更深入地解析野生鱼类与养殖鱼类间的相互作用机制——此类相互作用对疾病传播、养殖鱼类应激反应、野生鱼类行为模式及营养状态等诸多方面均具有广泛影响。
精准监测野生鱼类的存在状况及其时空分布变化规律，将有助于我们进一步明晰这类相互作用的动态过程及其后续影响。
当前已有诸多研究尝试借助人工智能（artificial intelligence）识别鱼类物种，但可用于训练神经网络（neural networks）的公开图像数据集仍较为稀缺，而涵盖水产养殖相关物种的数据集更是寥寥无几。
为此，我们公开了一套面向深度学习（deep learning）鱼类物种识别任务的标注图像（annotated images）数据集。该数据集包含9487张养殖鲑科鱼类（salmonids）的标注图像，以及3027张狭鳕（saithe）的标注图像，且预计未来数据集规模将持续扩充。
本数据集源自对挪威多家养鱼场网箱内外采集的近50小时视频素材的处理工作。
研究人员通过一套半自动系统对水下鱼类视频素材进行处理，以构建大规模图像数据集。该系统在迭代流程中融合图像处理技术与深度神经网络技术，实现对视频源中图像的提取、标记与标注。
该系统的详细技术细节已撰写为期刊论文，目前正处于同行评审阶段，待论文正式发表后，本数据集的相关信息将予以更新。