search

yosungho / LineTR

Line as a Visual Sentence: Context-aware Line Descriptor for Visual Localization (Line Transformer)
Other
233 stars 35 forks source link

About the number of network loops #22

Open yiyihao2000 opened 5 months ago

yiyihao2000 commented 5 months ago

Your work is excellent, I would like to ask a question, may I ask you, I see that the number of layers of the network in the paper is 12, why is the n_line_descriptive_layers=1 in the code, is the part of the network circulating 12 times elsewhere?

yosungho commented 5 months ago

Hello, it turned out that I used n_line_descriptive_layers = 1 for the paper result. I had a chance to train the network with a greater number of layers and it helped to enhance matching performance.

yiyihao2000 commented 5 months ago

您好，原来我用 n_line_descriptive_layers = 1 作为论文结果。我有机会用更多的层来训练网络，这有助于提高匹配性能。

You mentioned in your paper that the lineTransformer network has 12 layers, where is this reflected in the code, maybe my negligence caused me not to find it, please give me a hint, thank you very much

yosungho commented 5 months ago

I used only a single 1 layer even when I got the paper result. The paper should be corrected to be 1 layer.