Which design pattern suits well for an application that has to:
- load a large temporal data for example timeseries, video, point cloud frames
- process individual datapoints or frames (for example video object tracking)
- output a visualization of results
- save results in a compact form for further analysis or re-visualization that skips first 2 steps
Its clear that there should be some sort of buffering and parallel processing. It is also clear that most of the video editing software are based on the same principles, in other word there should be a design pattern for this type of software. Is there a well defined software design pattern for such applications, and how is it called?
There is no universal design pattern for processing large amounts of temporal and video data. You correctly stated that there are requirements for such a system, and there are software platforms that meet these requirements. Multiple design patterns are used to implement these platforms, but I believe there is no "silver bullet".
For example, for analysis of structured temporal data you could use an ELK stack (Elastic Search, Log Stash, Kibana). This is a widely accepted solution, that supports parallel processing, parallel storage, circular input buffering, inverse indexing and other architectural patterns.
There must be platforms for video processing, supporting on-the-fly object identification, facial recognition, trajectory estimation, etc. These are not my field of expertise, though.
Temporal data processing application depends on subdividing the load among the participated CPUs making number of jobs to be equal to number of CPUs. and these CPUs will work as pipeline structure.
Hi,
have you already studied this?
https://martinfowler.com/eaaDev/timeNarrative.html
B.R.,
Ari
- Badges
- Science topic
- Similar topics
- Methodology
- Research Methodology
- Research Papers