Attaching the timestamp to each trash, and not to each box

[bertrandlalo]

etl/etl/utils/ai.py

Lines 138 to 147 in 159be3e

	detected_trashes.loc[:, "box"] = np.apply_along_axis(
	lambda frame_to_box: str(list(frame_to_box[0].values())[0]),
	axis=1,
	arr=detected_trashes[["frame_to_box"]].values,
	)
	detected_trashes.loc[:, "frame"] = np.apply_along_axis(
	lambda frame_to_box: int(list(frame_to_box[0].keys())[0]),
	axis=1,
	arr=detected_trashes[["frame_to_box"]].values,
	)

See discussion with Charles here:
you may (and it happens often) have a single detected trash that is found in mutliple consecutive frames (1 trash = several boxes), for instance:

{'detected_trash': [
                    {'frame_to_box': {'2': [0.43, 0.44, 0.49, 0.5], '3':[0.42, 0.44, 0.48, 0.5]},
                       'id': 0,
                       'label': 'fragments'}
}

I suggest attaching the timestamp to each trash, and not to each box. The timestamp of the trash should then be the average of the timestamps of the frames with boxes detected (in the example above, (timestamp of frame 2 + timestamp of frame 3)/2).

surfriderfoundationeurope/mot#35

[bertrandlalo]

This should fix it :

62c1c75

I suggest two ways of identifying the more relevant frame for the trash to be chosen as a reference :

• either the one where box area is the larger (assuming the trash hence is nearer from the kayak's location)
• or (what we discussed yesterday), the frame in the middle (in time).
  What do you think @AntoineBruge , @cl3m3nt ?

Actually, I thought of this alternative because I was wondering what we should do if we have frame 1, 2, 3, 24, 25 (for example, user turns gopro around or kayak goes back and forth..)
I may go have a look on how MOT handles the tracking to have my answer :)

[cl3m3nt]

Reviewed