How can I run the iteration process in virtual scanning mode?

[inhwanlee]

In your paper, it says, "In practice, we terminate the scanning once the overall confidence does not improve any more. That is, the acquisition completes when the global confidence difference between two consecutive iterations is below a threshold (0.005 by default)."

Is there any code for running this process? Automatically compute NBVs and merges scanned points and compute NBVs again without pressing any button until overall confidence does not improve anymore.

Thank you. Inhwan Lee

[inhwanlee]

How did you evaluate the overall confidence for each NBV calculation loop?

[wsh312]

Hello Inhwanlee,

Thanks for asking! I don't remember we did much work on the stopping criteria. I think we perform SICP for the new scans and compute the matching error. If the error is too big, we don't use that scan and if the error is too small, these scans don't contribute much to the overall confidence. Something along this line. And I agree this is a difficult problem in practice.

Maybe you can check the "OneKey..." function in the code and see if we did something meaningful there. Thanks.

Best, Shihao

inhwanlee notifications@github.com 于2018年8月24日周五 上午6:18写道：

How did you evaluate the overall confidence for each NBV calculation loop?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/sunwaylive/quality-driven-poisson-guided-autoscanning/issues/9#issuecomment-415646464, or mute the thread https://github.com/notifications/unsubscribe-auth/AEcYpVC5BBWijTXEBQy6lTXkeEJkcVz2ks5uT36KgaJpZM4WHvFT .

[inhwanlee]

Thank you for answering to my questions. I'm sorry, but I have further question about the code.

In the code there is hole confidence which is not mentioned in your journal paper. How is the performance of hole confidence compared to smoothness&completeness confidence?

Sincerely, Inhwan Lee

[wsh312]

Hi Inhwan,

Maybe I'm wrong, but I guess they are just the same thing of a different name?

Best, Shihao

inhwanlee notifications@github.com 于2018年9月6日周四 下午1:58写道：

Thank you for answering to my questions. I'm sorry, but I have further question about the code.

In the code there is hole confidence which is not mentioned in your journal paper. How is the performance of hole confidence compared to smoothness&completeness confidence?

Sincerely, Inhwan Lee

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/sunwaylive/quality-driven-poisson-guided-autoscanning/issues/9#issuecomment-419066960, or mute the thread https://github.com/notifications/unsubscribe-auth/AEcYpdQfSRNOlbuJZlUvbSGzMcnPRMYJks5uYQ3UgaJpZM4WHvFT .

[inhwanlee]

In your code, hole confidence measures whether there are many nearby scanned data or not. Meanwhile, smoothness&completeness confidence algorithm which is mentioned in your Siggraph journal measures smoothness and completeness in more sophisticated way.

However, in your commits on 18 Aug 2014, you said "add hole confidence, which improve the result a lot". which seems hole confidence algorithm is a lot more simpler and not sophisticated and different from your method in Siggraph journal.

You added hole confidence code later, so I'm curious why you added this "Hole confidence" measure. Did you publish any journal about this "Hole confidence" measure in autoscanning?

Sincerely, Inhwan Lee

[wsh312]

Sorry, I can't recall anything significant on this one. Maybe it is related to the section 5.2 or 5.3 in the paper? According to your description, I think it makes sense to avoid scanning the same region over and over again (some hole just can't be physically scanned because of material for example).

inhwanlee notifications@github.com 于2018年9月13日周四 上午10:13写道：

In your code, hole confidence measures whether there are many nearby scanned data or not. Meanwhile, smoothness&completeness confidence algorithm which is mentioned in your Siggraph journal measures smoothness and completeness in more sophisticated way.

However, in your commits on 18 Aug 2014, you said "add hole confidence, which improve the result a lot". which seems hole confidence algorithm is a lot more simpler and not sophisticated and different from your method in Siggraph journal.

You added hole confidence code later, so I'm curious why you added this "Hole confidence" measure. Did you publish any journal about this "Hole confidence" measure in autoscanning?

Sincerely, Inhwan Lee

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/sunwaylive/quality-driven-poisson-guided-autoscanning/issues/9#issuecomment-420922435, or mute the thread https://github.com/notifications/unsubscribe-auth/AEcYpVqIIY6WflpdGxFf0gUXbHzVUnv1ks5uahO2gaJpZM4WHvFT .