action #14960
closedGSoC 2017
0%
Description
I would like to get openQA into next year GSoC and in this issue I would like to gather ideas and related discussions to prepare us before official beginning. Also note that personally I have no experience with GSoC itself so any help is appreciated.
Official timeline is here. Important events for mentoring org applications are:
January 19 16:00 UTC Mentoring organizations can begin submitting applications to Google
February 9 16:00 UTC Mentoring organization application deadline
Consider this as a brainstorming list, thus don't hesitate to add any silly idea. Just keep in mind that the task should be something one can work 2-3 months on it.
Once we agree on some solid idea I propose to create separate issue for that one (maybe new category 'gsoc' can be created) and linked back here.
Since the ideas can get lost in discussion, please update this top post list:
Ideas
- GPU accelerated needle matching, video encoding
- NN based classification of the SUT output - goal: new branding doesn't affect test results
Updated by coolo over 7 years ago
Please consider the following side rules for GSOC: These are students, mostly junior programmers. The mentor needs to understand the problem very well to give guidance, but it should still be big enough to have a student run for a period of time without supervision.
You have to assume that the project will not finish or will not be successful - and what is most often granted: that the student will not maintain his code. So every complicated change in our core will need a lot of documentation, which eats up the time the student has.
So projects most fitted to GSOC are applications around the core. Tasks that will benefit the system, but if missing won't harm. So ideas (just shooting out of the hip):
- live view transferring deltas ("VNC")
- calculate suggestions on environment variables to needles (cluster analysis on needle matches and failures)
Updated by coolo over 7 years ago
btw: I'm not sure what your university does with NNs, but to come up with a NN design for QA will bring you a nobel price :)
Updated by oholecek over 7 years ago
I don't mean analysing the test results, but to create and train NN which will be able to classify what is on the SUT screen - plain desktop screen, bootloader, application window etc. and compare only interesting details to expected result. So ideally when i.e. branding changes this NN will still be able to match previous needles.
Updated by oholecek about 7 years ago
- Status changed from New to Rejected
Hmm... I missed this and failed to to do anything relevant for this year application.