蚂蚁集团推出20亿参数多模态遥感基础模型SkySense
近日，蚂蚁集团旗下的人工智能平台蚂蚁百灵大模型团队宣布推出20亿参数多模态遥感基础模型SkySense。该模型在世界计算机视觉顶会CVPR 2024上获得接收，标志着蚂蚁百灵大模型在多模态领域取得了新的突破。
SkySense是迄今为止国际上参数规模最大、覆盖任务最全、识别精度最高的多模态遥感基础模型。在17项测试场景中，其指标均超过国际同类产品。
SkySense集成了遥感图像分类、目标检测、语义分割、图像生成等多项遥感任务，可广泛应用于地貌、农作物观测和解译等领域，有效辅助农业生产和经营。
据了解，SkySense模型的训练使用了来自全球多个地区的遥感卫星图像和标签数据，总量超过100TB。模型采用先进的深度学习算法和自监督学习技术，能够从海量遥感数据中学习丰富的特征表示，并实现对不同遥感任务的泛化。
蚂蚁集团相关负责人表示，SkySense模型的推出将为遥感领域的研究和应用提供新的动力。未来，蚂蚁百灵大模型团队将继续探索遥感大模型的更多可能性，推动遥感技术在农业、环境监测、城市规划等领域的广泛应用。
SkySense模型已开源，研究人员和开发者可通过蚂蚁百灵大模型平台获取模型和相关资源，进行进一步的研究和开发。
英语如下：
**Headline: Ant Blazing Releases Massive Remote Sensing Model to Empower Agricultural Production**
**Keywords: Remote sensing model, Multi-modal, Agricultural applications**
**News Content:** Ant Group’s artificial intelligence platform, Ant Blazing,has announced the release of SkySense, a 2-billion-parameter multi-modal remote sensing foundation model. The model was accepted at the top computer vision conference, CVPR 2024, marking a new breakthrough for Ant Blazing in the multi-modal domain.
SkySense is the world’s largest multi-modal remote sensing foundation model to date, with the most comprehensive task coverage and the highest recognition accuracy. It outperforms international peers in 17 test scenarios.
SkySense integrates multiple remote sensing tasks, including remote sensing image classification, object detection, semantic segmentation, and image generation. It can be widely applied in areas such as landform, crop observation, and interpretation, effectively assisting agricultural production and management.
The training of the SkySense model utilized remote sensing satellite images and label data from multiple regions around the world, totaling over 100TB. The model employs advanced deep learning algorithms and self-supervisedlearning techniques, enabling it to learn rich feature representations from massive remote sensing data and generalize to different remote sensing tasks.
Ant Group’s relevant负责人stated that the release of the SkySense model will provide new impetus for research and applications in the field of remote sensing. In the future, the Ant Blazing team will continue to explore the potential of remote sensing large models and promote the wide application of remote sensing technology in agriculture, environmental monitoring, urban planning, and other fields.
The SkySense model has been open-sourced. Researchers and developers can obtain the model and related resources through the Ant Blazing platform for further research and development.
【来源】https://www.geekpark.net/news/331706
Views: 1