Yang Gao

> Hi, thanks for your injterest. This file is generated from the YFCC100M dataset, "K" means 3x3 intrinsic matrix and "R" means relative rotation matrix. Evaluation metrics like the AUC...

> 您好，很感谢您提供的项目！我想在superponit和superglue的预训练模型上微调自己的数据集。我使'--keypoint_encoder'=[32,64,128, 256]，'--superpoint_weights'="superpoint/models/weights/superPointNet_coco_descriptor_256.pth.tar"， '--pretrain_weights'="indoor"，输入指令：python superpoint_glue_train.py --descriptor_dim 256 --sinkhorn_iterations 30。运行起来函数的Loss，请问我的操作有问题吗？ 不好意思少打了一些内容，我运行起来的Loss开始为inf，后面为Nan,请问您知道为什么吗？我的操作有问题吗？

hello,had you solve this problem？

> @Skydes Thanks for your suggestion, I have used rootsift. Although the loss decreases and the matching performance get better, the loss decreases to 1.0 and it converges. Does sift(DOG)...

Anyone had solved it? thanks! help!

> Hi - I also was interested in this. > > Maybe there is a better way but I was able to recover the depth map by multiplying the `ray_pts`...

> Hi, @sunset1995 How can I caculate the ground depth in real-world scale in the [dmpigo.py](https://github.com/sunset1995/DirectVoxGO/blob/341e1fc4e96efff146d42cd6f31b8199a3e536f7/lib/dmpigo.py#L333). I found the render depth is between 0 and 1.  > > I...

> Thanks for your repo! There comes an issue when I train my own dataset with nerf_w. It seems like the torch has something wrong, but I don't have an...

> Yes. You just need to hack some codes in `Tracking::GrabImageMonocular()` and overload `Frame` constructor with `GCNextractor`. Can you tell the details?

> Yes. You just need to hack some codes in `Tracking::GrabImageMonocular()` and overload `Frame` constructor with `GCNextractor`. Can you tell the details?