Hello professor, your work has really enlighten me. However, I have some questions. In your paper 4.1 chapter, you write "discretize the distribution by sampling depth hypotheses". Could you please tell me what does that mean or how did that? Thank you!

evaluate.py: error: unrecognized arguments: --checkpoint_path /home/user/fyy/deltar/checkpoints/nyu.pt
这条指令是导入训练的权重吗，为什么是无效的呢

Hi ,
Thanks for the great work! I'm curious how to access the histogram data from the L5 sensor? I really appreciate you can share your capture system:)

Regards.

大佬您好，我看到你的数据增强部分好像只有Totensor?其余的数据增强部分是没有加吗

Thank you for the nice work. The evaluation and demo with your weights and configuration works like in the paper. I have installed the library versions provided in the requirements.txt

My problem now is that I want to adapt, e.g. shrink, it to fit my purpose therefore I have to train my own models. So I tried to train the entire DELTAR model as it is provided by you. I use 25k nyu images. I have tweaked the code here and there to fit my data and I changed it with the two additional remarks from below. In this setting I trained it for 50 epochs, but I have also trained for 25 epochs with the settings provided in the configs for nyu. I then get the following results for validation:

and training loss:

images from the training set looks good though:

but it does not generalize that well:

My questions now are:

• How do you load/preprocess the nyu dataset?
  I used this python version of the toolbox: https://github.com/Lucass97/nyuv2_raw_dataset_extractor
• Do you have other suggestion what to watch out for?

Additional remarks:

• I found that the pointNet parameters where not trained in the network before as they are not added to the parameters for the optimizer in

    def get_10x_lr_params(self):  # lr learning rate
        modules = [self.decoder, self.depth_head, self.conv_out]

• In the nyu dataloader in "train" mode the image should be normalized t0 [0,1] (image = np.array(image, dtype=np.float32) / 255.0) before the random_crop(...) and train_preprocess(...) augmentation as it is clipped there to [0,1]

How did you do on Interpolation for solving misalignment? Could you please describe some more details or some code snippets?

hello,
first, thanks for sharing your work

i attempted to reproduce the same result as the example image from readme.md
but there is a big difference in quality

i used the provided weight, data, and config
should i use different parameters to reproduce the result?

i attatched the result that i made

please understand that the file is compressed due to file size issue

大佬，你的工作真的太有意义了。我有个疑问，我看了论文注意到最后是在ZJU-L5数据上面进行fine-tune，这个数据量是非常小的，总共才1000多张吧，这么小的数据量fine-tune的时候不会过拟合吗？因为我目前也做了单目深度估计模型的fine-tune，我混合了几个公开数据大概20多万张，训练了一个模型，然后加载我训练好的模型在我自己采集的2000多张数据上面fine-tune的时候过拟合了。我不清楚是由于数据量太小过拟合了，还是其他什么原因。如果可以的话，可以多说一些fine-tune的细节吗？比如一些fine-tune的示例代码。

大佬您好，我发现训练时候模拟的tof数据是像素相对于相机平面的距离，但是真实tof传出来的数据是像素相对于相机中心的距离，您在处理tof数据的时候是将原始tof数据转换成了相对于相机平面的距离吗？

您好，我想问一下tof输入是否有归一化？好像看代码依旧是真实深度值？

感谢您的工作！
我在实验过程中发现，构建直方图的部分使用了参数--simu_max_distance=4，但是--max_depth=10。
想问一下为什么要这样设置，以及这两个参数的关系。

Hello，your work is excellent and inspiring. I want to build a system to build dataset likes yours. Will you released the code of compute the relative rotation and translation between the color camera and the L5 sensor? thanks

感谢大佬的工作！
项目主页中的补充材料下载链接失效了，请问作者可以提供一个新的下载链接吗？
非常感谢！！

conda create --name deltar --file requirements.txt does not work for me as it does not find a lot of packages. PackagesNotFoundError: The following packages are not available from current channels:
I installed conda on my wsl with Ubuntu-22.04.
Which conda channels do you use?

Hi Authors, Awesome paper, I have pretty much the question said in the subject. I have a ToF sensor which has a depth image resolution of 224x172? Can I retrain the algorithm for this sensor and use it for RGB-D mapping while this algorithm gives super resolved depth image in real time?

大佬，你的工作太有意义了，请问你什么时候开源代码呢？

Hi,
Congratulations on the amazing work! When do you plan to release the code in public?

大佬的工作很棒！我想请教一下怎么获取论文中所述选出的数据集（可以在kaggle上获得吗），我想要进一步搭建自己的数据集以供训练，谢谢！！！

你好，我看到您的工作训练的时候，是使用RGBD数据来仿真TOF数据，从而进行训练的。我想问一下，请问从rgbd到tof具体是怎么得到的呢？感谢您的帮助。

另外，我想问一下，您的Deltar算法可以扩展到其他TOF数据嘛？比如说3040这样的，而不是88的？需要做哪些改动呢？

谢谢，期待您的回复