thusodangersimon / image_diff Goto Github PK

The parallelization is not yet doing what it is meant to! Fixing it right now...

The last few frames of the video are creating NANs need to sort this out...

I think the main bottleneck will be reading the video! So if we read it once in make_master_im we might have to try not to read it again in image_chi...

The code as it is you can change the different frame numbers!

For example, with 30 frames it takes 1 min while 3000 frames it take 5 min!

We may need to think a little bit more on this, because if we find that we need higher values for the background we'll be running into "trouble"! But I mean, things are still reasonable...

We will need an output file for the chisquared -- this is then also used in the plots!

something like this: http://arxiv.org/abs/1207.5578
or http://arxiv.org/pdf/0710.3742v1.pdf
http://www.variation.com/cpa/tech/changepoint.html

I'm getting this error when using the code:

out.append(temp_image[sample])

For some reason, at the moment with the new code parallelised this seems to be an issue with the very last array! But when I remove [sample] it appears to work.

I will check this tomorrow when I'm in because I can't see what the effects are to the rest of the code...

• Need to update the timezone code
• Perhaps have a switch for the user to choose whether or not to include a timezone

This will aid users to follow how to use the code

Need to create a whole load of images that will then be appended to the output video so that we tracks where in the video an image comes from.

Insert a small pause when zipping the files because in some cases when writing out the reduced chisquared file results in sometimes a file starting on top of the other! We can play with the value below later, specially when we have more cores!

time.sleep(0.2*self.rank)

We need to output one jpg image with all the reduced chisquared

We should sort the file with the frames and reduced chisquared before it goes into the plot area! This should be as simple as inserting:

self.frame_no.sort()

I think this will also automatically bin the self.frame_chi into the right place!