Democratizing Deep Learning Applications in Earth and Climate Sciences on the Web: EarthAIHub

Most deep learning application studies have limited accessibility and reproducibility for researchers and students in many domains, especially in earth and climate sciences. In order to provide a step towards improving accessibility to deep learning models in such disciplines, this study presents a community-driven framework and repository, EarthAIHub, that is powered by TensorFlow.js, where deep learning models can be tested and run without extensive technical knowledge. In order to achieve this, we present a configuration data specification to form a middleware, an abstraction layer, between the framework and deep learning models. Once an e...  more

Most deep learning application studies have limited accessibility and reproducibility for researchers and students in many domains, especially in earth and climate sciences. In order to provide a step towards improving accessibility to deep learning models in such disciplines, this study presents a community-driven framework and repository, EarthAIHub, that is powered by TensorFlow.js, where deep learning models can be tested and run without extensive technical knowledge. In order to achieve this, we present a configuration data specification to form a middleware, an abstraction layer, between the framework and deep learning models. Once an easy-to-create configuration file is generated for a model by the user, EarthAIHub seamlessly makes the model publicly available for testing and access using a web platform. The platform and community-enabled model repository will benefit students and researchers who are new to the deep learning domain by enabling them to access and test existing models in the community with their datasets, and researchers to share their novel deep learning models with the community. The platform will help researchers test models before adapting them to their research and learn about the model details and performance.