Why does blend_images take so long to run?

[bakermanbrian]

I've been running this project and it's very fast (thanks!) except for this one part:

    def blend_images(self, imgs, masks, corners):
        for idx, (img, mask, corner) in enumerate(zip(imgs, masks, corners)):
            if self.timelapser.do_timelapse:
                self.timelapser.process_and_save_frame(
                    self.images.names[idx], img, corner
                )
            else:
                self.blender.feed(img, mask, corner)

I have self.timelapser.do_timelapse as False and I have done time tests and the self.blender.feed(img, mask, corner) line is quite fast. It seems that the loop iteration is the slow part.

So even this is slow for me:

    def blend_images(self, imgs, masks, corners):
        for idx, (img, mask, corner) in enumerate(zip(imgs, masks, corners)):
            pass

How would I go about speeding up this loop iteration?

[lukasalexanderweber]

this is because stitching uses generators to optimize memory usage. Whenever yield is used, a generator is used. Elements are only loaded when they are actually needed. This is the case in the corresponding blend_images function at the end of the pipeline, even if nothing is done with the data (pass). To see which part of the stitching takes the longest, it would probably be better to measure the times in verbose stitching. whenever list() is called, a generator is converted to a list, which means that the data is actually loaded. If you are interested in the topic in more detail, you can also take a look at the performance comarison test.