Add VOC format

[MattKleinsmith]

A fresh pull request with respect to: https://github.com/Cartucho/yolo-boundingbox-labeler-GUI/pull/6

I made a fresh pull request so that only my changes are highlighted.

If a user tries to use multiple bounding box formats on the same image, there will be an error. I added error messages to guide them.

---

YOLO is the default and python run.py will still work normally.

To enable VOC: python run.py --format voc

Users will likely have to further process the bounding box text files to match their pipeline, but this should make it easier for them, since they won't have to undo YOLO normalization.

[MattKleinsmith]

Some changes I will propose after this pull request (the code is ready):

• Making colors consistent between labeling sessions ("I got used to blue meaning one thing, but now it means another.")
• Adding the option to make the bounding box form upon mouse-up, instead of requiring two clicks (the former is conventional GUI dragging behavior and requires less clicking) (I might need to update it for the double-click feature you added)
• Adding the filename to the help overlay to make debugging easier
• Adding the option to display images in order
• Adding the option to start the session at the most recently labeled image

I mention these so we don't duplicate work and so we don't accidentally make mutually exclusive changes.

[Cartucho]

@MattKleinsmith great ideas!

[Cartucho]

@MattKleinsmith this should be changed: line 91 and 92

def voc_format(class_index, point_1, point_2):

Order: xmin ymin xmax ymax class

# Top left pixel is (1, 1) in VOC

From:

• xmin, ymin = point_1[0] + 1, point_1[1] + 1
• xmax, ymax = point_2[0] + 1, point_2[1] + 1

To:

• xmin, ymin = min(point_1[0], point_2[0]) + 1, min(point_1[1], point_2[1]) + 1
• xmax, ymax = max(point_1[0], point_2[0]) + 1, max(point_1[1], point_2[1]) + 1

This way we make sure we store xmin, ymin, xmax, ymax