anchors in your code

[KennithLi]

Hi, thank you for your work! I want to know how you get your anchors for the datasets? Could you share the code or anything else? I did not find any detailed explainations in your paper. Only a little information is shown in model/mcn_model.py line 451:

anchor_mask = [[6,7,8], [3,4,5], [0,1,2]] if num_layers==3 else [[0,1,2]] ##due to deleting 2 scales  change [[6,7,8], [3,4,5], [0,1,2]] to [[0,1,2]]

But the data in your data/yolo_anchors.txt is

137,256, 248,272, 386,271

for [[0,1,2]], deleting [[6,7,8], [3,4,5] scales. I want to get the whole anchors for the datasets.

Hearing from you, thanks~

[luogen1996]

Hi, thank you for your work! I want to know how you get your anchors for the datasets? Could you share the code or anything else? I did not find any detailed explainations in your paper. Only a little information is shown in model/mcn_model.py line 451:

anchor_mask = [[6,7,8], [3,4,5], [0,1,2]] if num_layers==3 else [[0,1,2]] ##due to deleting 2 scales  change [[6,7,8], [3,4,5], [0,1,2]] to [[0,1,2]]

But the data in your data/yolo_anchors.txt is

137,256, 248,272, 386,271

for [[0,1,2]], deleting [[6,7,8], [3,4,5] scales. I want to get the whole anchors for the datasets.

Hearing from you, thanks~

Thank you for your interest. The most of objects in Ref-series are middle or large size, so we only keep one detection branch. And the corresponding anchor is set to 137,256, 248,272, 386,271. If you want to use more scales/anchors, you can utilize kmeans to calculate appropriate anchors for your datasets.