Ilya Basharov

Lol, NaN's solved by change this [line](https://github.com/SuLvXiangXin/zipnerf-pytorch/blob/a086113b219901844c90bd30301e4e1a1ffa7b19/internal/models.py#L439) ```python3 weights = torch.erf(1 / torch.sqrt(8 * stds[..., None] ** 2 * self.encoder.grid_sizes ** 2)) ``` to ```python3 weights = torch.erf(1 / (8...

@SuLvXiangXin I successfully did pose refinement via mip-nerf 360 scene contraction implemented [here](https://github.com/nerfstudio-project/nerfstudio/blob/c7def8ce8cb42fc2869a31ec886e5797e076489a/nerfstudio/field_components/spatial_distortions.py#L42). Basically now I try to analyse your contraction function (still nans in gradients). Also I have a...

@SuLvXiangXin I've solved nan gradients on contract function. see my PR to nerfstudio https://github.com/nerfstudio-project/nerfstudio/pull/2242. the idea is ```python3 ... # prevent negative root computations clamped_mag = mag.clamp_min(1.0) det_13 = (torch.pow(2...

@datasciritwik small value like 1e-8

@aash1999 Hello! Thanks for your work! Did you evaluate the results? What about metrics?

> Hi @Ilyabasharov I implemented as mentioned by the paper : https://arxiv.org/pdf/2309.16393.pdf You can find the results as mentioned in them. And also I am training for the same to...

@jameslahm do you plan to implement yolov10-p2 version ?

@shyhyawJou see paper https://arxiv.org/pdf/2112.00322 for Mobius parametrisation for improving OBB detection to handle breaks at 0/360, or 0/2pi radians. Here is the code https://github.com/SamsungLabs/fcaf3d/blob/bcc7ddc81651b3d586caf4990bbfc962cc77de53/mmdet3d/models/dense_heads/vote_head_v2.py#L224