search

cherubicXN / hawp

Holistically-Attracted Wireframe Parsing [TPAMI'23] & [CVPR' 20]
MIT License
298 stars 52 forks source link

Questions about the paper and code version #47

Closed eleveneee closed 1 year ago

eleveneee commented 1 year ago

Thanks for your work! That‘s excellent! I am very interested in your research. When I learned about you, I found that there are several versions of the paper and code. I want to ask about the relationship between them. Q1：Is《Holistically-Attracted Wireframe Parsing: From Supervised to Self-Supervised Learning》 an improvement of 《Holistically-Attracted Wireframe Parsing》？ Q2：If I want to run the code, which one is better to learn? Is it the latest version? But I haven't seen the relevant running tutorials in the new version. Do I still need to refer to the previous code links?

cherubicXN commented 1 year ago

A1: Yes, it is a journal extension of our conference paper. A2: It depends on your requirement. The previous version is tagged as 0.1 in my repo, you can also visit it from https://github.com/cherubicXN/hawp/tree/0.1. If you want to use deep models like the traditional LSD without tuning, HAWPv3 would be better. If you need to compare your method with ours in a fully-supervised learning paradigm, both HAWPv1 and HAWPv2 are ok.

cherubicXN commented 1 year ago

Because the extended version is a preprint, you can use the old code without considering the updates. Of course, if our updates are useful for you, I will be very happy!

eleveneee commented 1 year ago

Thank you very much for your timely reply! Very useful. Good luck!