The structure of Lotus-D

[Cuirundi]

In the training of Lotus-D, is only the image used as input to the UNet, or is the image concatenated with the ground truth depth as input?

[jingheya]

Sorry for my late reply. Only the image latent is used as input to the UNet model (4 channels).

[Cuirundi]

Dear Jingle He: Thank you for you reply. There is a question that I still want to consult you. Should image reconstruction and depth estimation be carried out alternately for each batch during training,such as doing depth estimation for one batch and image reconstruction for the next batch.Or should the two tasks be carried out separately in two unet processes for one batch? Thank you again for your reply. Looking forward to your next reply. ------------------ 原始邮件 ------------------ 发件人: "EnVision-Research/Lotus" ***@***.***>; 发送时间: 2024年12月9日(星期一) 晚上9:30 ***@***.***>; ***@***.******@***.***>; 主题: Re: [EnVision-Research/Lotus] The structure of Lotus-D (Issue #25) Sorry for my late reply. Only the image latent is used as input to the UNet model (4 channels). — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>