Open Rienzi0 opened 9 months ago
Sorry for the mistake. The outputs need to be multiplied with 10 to maintain a same range of the ground truth. We will modify the codes later. Thank you for noticing!
------------------ Original ------------------ From: Zihui Zhao @.> Date: Wed,Dec 6,2023 4:05 PM To: htx0601/AGG-Net @.> Cc: Subscribed @.***> Subject: Re: [htx0601/AGG-Net] Source code problem (Issue #4)
Hello, I find the depth images in NYUv2 dataset you applied ranged in [0,10], but the final layer of network is a sigmoid layer. How can this work when you expect an output of sigmoid function to fit a distribution in range[0,10]? Are there any other details in your codes which are not contained in this repository?
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: @.***>
Hello, I find the depth images in NYUv2 dataset you applied ranged in [0,10], but the final layer of network is a sigmoid layer. How can this work when you expect an output of sigmoid function to fit a distribution in range[0,10]? Are there any other details in your codes which are not contained in this repository?