Open cong-code opened 1 year ago
abrilcf
commented
1 year ago Hi, those parameters correspond the camera/scene. Near and far are the scene bounds (you can find more on this from NeRF's original repo), radius depends on these two. u and v are for azimuth and polar angle, which are set in ranges for a rough distribution of the camera poses. And fov is the field of view. You will find also useful the original graf repo for how these are set for different datasets.
cong-code
commented
1 year ago Dear author： I am very happy to receive your reply. The camera/scene parameters of graf correspond to the natural image, but there should be differences between the camera in X-ray imaging. For example, I think the x-ray imaging near and far are set to 7.5, 12.5 seems not suitable, which is equivalent to a human body thickness of only 5. I don’t know if I understand it correctly. Thanks and regards.
my name: Wengcong zhang @.***
------------------ 原始邮件 ------------------ 发件人: "abrilcf/mednerf" @.>; 发送时间: 2023年5月31日(星期三) 晚上11:35 @.>; @.**@.>; 主题: Re: [abrilcf/mednerf] Parameter (Issue #12)
Hi, those parameters correspond the camera/scene. Near and far are the scene bounds (you can find more on this from NeRF's original repo), radius depends on these two. u and v are for azimuth and polar angle, which are set in ranges for a rough distribution of the camera poses. And fov in the field of view. You will find also useful the original graf repo
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>
cong-code
commented
1 year ago Dear author： I am very happy to receive your reply. The camera/scene parameters of graf correspond to the natural image, but there should be differences between the camera in X-ray imaging. For example, I think the x-ray imaging near and far are set to 7.5, 12.5 seems not suitable, which is equivalent to a human body thickness of only 5. I don’t know if I understand it correctly. Moreover, In the test phase, when given a 0-degree anterior-posterior view, how to specify how to render and predict the lateral view, render_xray_G.py is a randomly generated pose parameter, and the desired pose cannot be specified. Thanks and regards.
my name: Wengcong zhang E-mail: @.***
------------------ 原始邮件 ------------------ 发件人: "abrilcf/mednerf" @.>; 发送时间: 2023年5月31日(星期三) 晚上11:35 @.>; @.**@.>; 主题: Re: [abrilcf/mednerf] Parameter (Issue #12)
Hi, those parameters correspond the camera/scene. Near and far are the scene bounds (you can find more on this from NeRF's original repo), radius depends on these two. u and v are for azimuth and polar angle, which are set in ranges for a rough distribution of the camera poses. And fov in the field of view. You will find also useful the original graf repo
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>
Hi, author, I'd like to consult a few hyperparameters given in the code：radius, near, far, fov, umin, umax, vmin, vmax, and what they mean. How do you determine these for different X-rays devices。(u, v) denotes the center of patch? I understand that (u, v) is the central coordinate of patch, why is the Angle given.