STAY UP TO DATE

ImJoy: a new computational platform for deep learning

Researchers from SciLifeLab have developed a new online platform to address the pressing need for publically available deep learning bioinformatic tools – tools that can learn how to make correct decisions by themselves based on large data sets.

The new platform, known as ImJoy, facilitates the development and sharing of advanced data analysis tools, especially AI-based solutions for a broad range of biomedical research.

Deep learning becomes more popular every day but also requires a much more powerful software infrastructure, including stronger GPU:s or complex software environments, which makes easy-to-use tools difficult to build.  

To address the problem, researchers co-led by Emma Lundberg (SciLifeLab/KTH), developed ImJoy in collaboration with Institut Pasteur. The platform works with different operating systems and computing environments including mobile phones, workstations, cloud servers and institutional computing clusters. 

Developers, working with Javascript, Python and other script languages, can easily use the platform to build interactive and user friendly applications which can run on either desktop computers or mobile phones.

The study, published in Nature Methods demonstrates a whole range of AI/deep learning based applications including image segmentation, skin lesion analysis, DNA-/RNA- binding probability prediction and protein localization classification.

“Many open-source deep learning tools such as Keras and PyTorch are designed to be used like playing LEGO blocks, by developers. But are hardly accessible to most regular users. ImJoy is a very thin layer that transforms deep learning model prototypes into software products for end users. We believe it will be a useful platform for the community.” says Wei Ouyang, lead author and a postdoctoral researcher in the Cell Profiling group.

ImJoy and its source code are freely accessible at https://imjoy.io


STAY UP TO DATE

Last updated: 2019-12-11

Content Responsible: admin(website@scilifelab.se)