the inputs of transient and appearance embedding
SensenZhang opened this issue · 4 comments
SensenZhang commented
Hello, I would like to know the input of transient_embedding and appearance_embedding. Thank you
rover-xingyu commented
Hi, the transient embeddings are randomly initialized and trained by optimizing the objective function, and the CNN encoder extracts the appearance embeddings.
SensenZhang commented
Thank you for your reply.
While debugging the code nerf-w, I was puzzled by the 'ts' in the following figure, which when initialized seems to be just a set of integers "img_num*[img_h*img_w]", independent of the characteristics of the image. I want to know the specific meaning of ts and how to obtain it.
Thank you very much.
谢谢您的回复。我在调试代码nerf-w的过程中，对下图中的 ‘ts’ 感到疑惑，'ts' 在初始化时似乎只是一组 “img_num*[img_h*img_w]” 的整数，与图像的特征并无关系。我想知道ts的具体含义，以及获取的方法。谢谢您！
…------------------ 原始邮件 ------------------
发件人: "rover-xingyu/Ha-NeRF" ***@***.***>;
发送时间: 2023年4月1日(星期六) 晚上7:11
***@***.***>;
***@***.******@***.***>;
主题: Re: [rover-xingyu/Ha-NeRF] the inputs of transient and appearance embedding (Issue #23)
Hi, the transient embeddings are randomly initialized and trained by optimizing the objective function, and the CNN encoder extracts the appearance embeddings.
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you authored the thread.Message ID: ***@***.***>
rover-xingyu commented
It denotes the id of the image.
SensenZhang commented
I get it. Thank you very much for your reply!
…------------------ 原始邮件 ------------------
发件人: "rover-xingyu/Ha-NeRF" ***@***.***>;
发送时间: 2023年4月1日(星期六) 晚上8:46
***@***.***>;
***@***.******@***.***>;
主题: Re: [rover-xingyu/Ha-NeRF] the inputs of transient and appearance embedding (Issue #23)
It denotes the id of the image.
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you authored the thread.Message ID: ***@***.***>