Annotation mean

[Aiden-0209]

I have two questions about labels in a dataset.

1. I don't know what the values for each file's label mean. DIVOTrack/datasets/DIVO/images/test/Shop_View1/det/det.txt Here's what's written in the file above. 1,-1,414.0,37.0,92.0,211.0,0.94

and then DIVOTrack/datasets/DIVO/labels_with_ids/test/Shop_View1/img1/Shop_View1_000001.txt Here's what's written in the file above. 0 231 0.592448 0.474537 0.057813 0.245370

I thought it was information about the frame, person id, and bounding box. But this doesn't seem to be accurate, so I'm asking.

2. Isn't the reason the folder names are separated for cross-view in the test set?

DIVOTrack/datasets/DIVO/labels_with_ids_cross_view/train DIVOTrack/datasets/DIVO/labels_with_ids_cross/test

[shengyuhao]

For Q1, "0 231 0.592448 0.474537 0.057813 0.245370" is frame id, person id, box_center_x/image width, box_center_y/image height, box width/image width, and box height/image width. (In the detection file, all person IDs are uniformly set to -1)

[shengyuhao]

For Q2, In "labels_with_ids_cross_view", starting IDs for different scenes are reset to 1, while "labels_with_ids_cross" has cumulative IDs, where the next scene starts from the previous scene's max ID+1. We will update them later.

[Aiden-0209]

Thank you for your answer :)