-
Hi, thank you for the great work!
I'd like to ask, did you train LightGlue on SuperPoint-Open's keypoints & descriptors? If so, are the weights available somewhere?
Or are the [LightGlue weights…
-
Asset/MagicLeap配下の証明書は使えませんでした。どうにも同階層に.privatekeyが必要のようです。
-
hi @fwilliams,
in recent pytorch versions it has been introduced [`torch.logsumexp`](https://pytorch.org/docs/stable/torch.html#torch.logsumexp) which I think that could be used to replace [`stabil…
-
In utils.py L292: https://github.com/magicleap/SuperGluePretrainedNetwork/blob/ddcf11f42e7e0732a0c4607648f9448ea8d73590/models/utils.py#L292
`f_mean = np.mean([K0[0, 0], K1[1, 1], K0[0, 0], K1[1, 1…
-
Hello, I saw the question you asked in magicleap/supegluepretrained, Could i refer to your train code in Scannet？ thank you！
-
Hi,
Superpoint from Magicleap seems to be particularly effective for detecting consistent features outdoors. According to Daniel DeTone (the author), it makes possible localization despite weather …
-
Starting in version 1.4.0 of PyTorch the default behavior of `torch.nn.functional.grid_sample` was changed to `align_corners=False`. This might affect the outputting descriptors for certain input imag…
-
Hi,
Superpoint from Magicleap seems to be particularly effective for detecting consistent features outdoors. According to Daniel DeTone (the author), it makes possible localization despite weather …
-
https://twitter.com/javierdavalos/status/1198016780809060352
-
At present, the shaders do not work properly with [Single Path Instanced Rendering](https://docs.unity3d.com/Manual/SinglePassInstancing.html), which is commonly used by XR headsets (HoloLens, MagicLe…