Message improvements (properties and methods).

[kkeroo]

This PR improves 3 messages:

1. Predictions msg - we add a property prediction for retrieving first prediction value. This is useful because Regression parser many times returns one value and to get the value we must type message.predictions[0].prediction. Now, we will just write message.prediction.
2. Classification msg - same idea as with Predictions msg. to get the most probable class and score we add two properties: top_class and top_score.
3. ImgDetectionExtended msg - two methods are added for easy conversion to [x1,y1,x2,y2] bbox and for [[x1,y1], [x2,y2],[x3,y3],[x4,y4]] rotated bbox (going from the top-left and clockwise)

[github-actions]

Test Results

125 tests   125 ✅  1s ⏱️
  1 suites    0 💤
  1 files      0 ❌

Results for commit c90782e.

♻️ This comment has been updated with latest results.

[github-actions]

☂️ Python Coverage

current status: ✅

Overall Coverage

Lines	Covered	Coverage	Threshold	Status
3757	1256	33%	0%	🟢

New Files

No new covered files...

Modified Files

File	Coverage	Status
depthai_nodes/ml/messages/classification.py	81%	🟢
depthai_nodes/ml/messages/img_detections.py	67%	🟢
depthai_nodes/ml/messages/prediction.py	87%	🟢
TOTAL	79%	🟢

updated for commit: c90782e by action🐍

[klemen1999]

LGTM

[jkbmrz]

LGTM. The added features seem useful! I'm wondering, though, if we are not making the messages too complicated for DAI to implement natively (e.g. for bbox-related methods, wouldn't it be much easier for DAI to have such transformations done on-host?)

[kkeroo]

LGTM. The added features seem useful! I'm wondering, though, if we are not making the messages too complicated for DAI to implement natively (e.g. for bbox-related methods, wouldn't it be much easier for DAI to have such transformations done on-host?)

Good question but IIRC Matevz said that we can add functions to msgs with no problem.