Possible Incorrect Foot Point Calculation in get_img_gt function

[visionNoob]

👋 Hi, I found a possible issue with the calculation of foot_pts in the get_img_gt function.

In the current implementation:

EarlyBird/EarlyBird/datasets/pedestrian_dataset.py

Lines 170 to 171 in fee7379

	foot_pts = np.stack(((xmin + xmax) / 2, ymin), axis=1)
	foot_pts = torch.tensor(foot_pts, dtype=torch.float32)

It seems that ymin is used as the foot position (Y-coordinate), but ymin typically represents the top of the bounding box (head position) rather than the bottom.

Wouldn't it make more sense to use ymax instead?
For example:

foot_pts = np.stack(((xmin + xmax) / 2, ymax), axis=1)
foot_pts = torch.tensor(foot_pts, dtype=torch.float32)

This way, foot_pts would represent the actual bottom of the bounding box, which is more aligned with the foot position.