The Media Collections Analysis resource from Advanced Research Computing (ARC) uses computing power to automatically detect objects, motions, audio or visual comparisons, image identification, generate captions across large media collections, among other techniques.
This resource is particularly useful when your project needs heavy computing power that analyzes large or many files, and too large to analyze by hand.
Purpose
There are many platforms on which you can run media collection analysis techniques, but if you have very large collections, it may be worth setting up your project on ARC hardware.
Benefits
Compute-heavy analysis tasks such as optical flow, object detection, or image captioning on very large media collections.
Challenges
There is some overhead time to get your environment set up on ARC systems.
Many of these techniques use software that do not have simple graphical interfaces. You may have to learn how to run these operations using, for example, Python scripting in Jupyter Notebook.
Workflow examples
- Optical flow motion correlation within and between video clips
- Object detection and counting in image collections
- Semantic captioning of images
- Voice transcription of audio recordings
Next steps
You will need an allocation on ARC for data storage (Chinook) and computation (Sockeye). Principal Investigators can apply for an allocation here. Other researchers can gain access to Chinook through their PI.
Costs
All ARC resources are free of charge for UBC researchers.
Contact
Contact ARC by email for specific questions or consultation at arc.support@ubc.ca.