qbonnard
commented
9 years ago Not sure whether we should discuss this by mail, here, in #72 or in #77 ;)
Just to be clear: we don't actually need video files to do the kind of tests you're talking about, right ? converting any video with avconv -i myvideo.mp4 mytestimage%d.jpg would produce something that is already usable in the current test suite. Not that it's a lot more complicated to open videos with opencv, but we wouldn't get differences among the decoding processes, and we can control timings more freely.
The idea behind the reprojection test was to have controlled, non artificial data for tracking/filtering too. We can have several, fixed tags displayed on a monitor, and a moving one in the middle. This allows to measure the relative movement when filming the monitor, and since this movement is controlled, we can compared it to the measured one. We can even use the ids of the tags to code the progress of the animation. To generate this animation, a simple svg+javascript should do. This even allows to replace chilitags by other tags, to compare.
We do not test chilitags works with tracking (as implemented in #38). To fix that, we need to develop video-based tests. One 'easy' way to do that is to shoot videos where a set of tag remain always in the field of view of the camera (but possible blurry, etc). We can then simply count on how many frames the tags were not all detected.