Contributing to Hydra

There are a number of ways in which you can contribute depending on your interests. Some of our interests are listed below. These may or may not match up with yours, but we are of course willing and open to collaboration to build a better Hydra.

System Upgrades

  • Deployment on Kubernetes

Database optimization / back end development

  • Data mining: Hydra’s databases are a wealth of information that can be mined to evaluate detector performance over time.

Data Science

  • Multimodal Hydra: Naturally, experiments have monitoring data in a multi-modal format. This includes images, numeric values, text data and so on.
  • Siamese Models: A Siamese model would be useful in Hydra if we wanted to compare an incoming monitoring image to a known, good reference image. The output of these types of models is a similarity score between 0 and 1.
  • One model for all images: It is possible to train a single model to look at all images.
  • Models to predict detector issues before they occur.
  • Instead of monitoring images, we could monitor the actual data.

Front end development

  • UI/UX improvements
  • Developing the interface for new data science approaches and integrations

Integrations

  • Pytorch: Currently, Hydra only requires Tensorflow based models.
  • MLFlow:
  • Logbooks: Direct submission from Hydra to experimental logbooks would avoid end users having to manage and clutter counting house computers with screenshots.