Multi-camera support

[aguscas]

It still needs to be done:

• Use the motion estimator when either the reference or the footage that is provided to the UI is a video. This is important since we may want to associate points from the reference and the footage from several different frames, and the camera(s) may have moved between those frames. DONE

• Next to "Remove", add button "Ignore" (to keep those buttons but not use them). You can later "Unignore" (?) them again if you want. DONE

• Add button to go to the beginning of the video in the UI, for either the reference or the footage. Would need to .release() the video and cv2.destroyAllWindows() before creating the new video instance, initialize a new Video instance with same path, re initialize the associated MotionEstimator, re define the "video", "motion_estimator", "motion_transformation", "current_frame" in the corresponding skipper dict, remove all drawing from the canvas and draw the new image. DONE

• Associate trackers (from different videos) that are close in the absolute frame. Can also use embeddings (might be useful) DONE

• Make demo using the UI to initialize the videos to a common frame. DONE

• Resize images: Maybe the user should be able to resize them from the UI, maybe give an option to resize both (reference and footage) independently. DONE

• 'save': pickle the transformation and the points variable. Should open a Save File Dialog DONE

• 'load': load a previous transformation and points variable to keep working on that transformation. Should be an Open File Dialog DONE

• Dockerize the demos DONE

• There is a bug in the UI whenever you try to use the file dialog more than once (segmentation fault: 11). Not much we I can do in this regard, it's a common problem with tkinter

• It would be nice to solve issue 306, so that we don't run into any problem if we process avi files (here, videos been synchronized is fundamental)

• Add gifs to the README