How does the normalization work?

[MoritzLennartz]

Hi,

Can you explain to me the normalization function in utils.predict.py?

When I use the sample image (python predict.py --images images/people-walking-on-pedestrian-lane-during-daytime-3.jpg) to print out the final keypoints (line 110 in predictor.py) used for prediction with the look model, one coordinate is greater than 1. How can this be?

Which coordinates exactly should the upper left corner and the lower right corner of the image have after normalization?

best regards
Moritz