mxnet_pose_for_AI_challenger
mxnet_pose_for_AI_challenger copied to clipboard
a problem about computing a person's center
“human_annotations”:若干长度为4的整数型数列,存储人体框的位置。其中前两个参数为人体框左上角点的坐标值,后两个参数为人体框右下角点的坐标值。
"human_annotations":contains the positions of human bounding boxes. The first two parameters indicate the top left coordinates of the human bounding box, and the last two parameters are the lower right coordinates.
person_center = [box[str(human)][0] + box[str(human)][2]/2, box[str(human)][1] + box[str(human)][3]/2 ] It is wrong to compute a person's center as this line code, it should be person_center = [(box[str(human)][0] + box[str(human)][2])/2, (box[str(human)][1] + box[str(human)][3])/2 ]
So is the the visibility flag: in coco, 1 visible, 2 non-visible, 0 means occluded by still on image , but in this dataset, 1 labeled and visible, 2 labeled but not visible, 3 not labeled so, the true relationship is: 3-0, 1-1, 2-2(this dataset-coco dataset)
@Marcovaldong Thank you. Those seem to bugs. @qqsh0214, do you agree?
@Marcovaldong Thank you. @dragonfly90 I also agree.
@Marcovaldong Could you make a pull request to fix the bug if you have some time?
@dragonfly90 OK, I'll pull a request soon.