Chenxu Zhou

> @rmx4 I let the scene_bbox encompass both the cameras and the objects. And when it ray sampling, i let it only samples within the object bounding box. I use...

> @dunbar12138 Hello, I have followed the steps mentioned in the README but imgs2poses.py does not generate colmap_depth.npy. Also, could you please share the training configuration for the DTU dataset...

> Hi, you can try this one to get full cameras' parameters. https://github.com/open-mmlab/mmdetection3d/blob/main/tools/dataset_converters/waymo_converter.py#L136 Can you share the scripts used to get full cameras' parameters? Thanks a lot.

> Admirable work! When I try to reconstruct a 198-frame waymo dynamic scene, the reconstruction effect is very poor. It seems that it can only succeed under 50 frames. Fine-tuning...

Thanks! How do you convert the raw .tfrecord waymo data to kitti format? I just use the sciprts provided by [Waymo_kitti_converter](https://github.com/caizhongang/waymo_kitti_converter), but the results can't be used directly.

> I want to know the specific location of the reconstructed point cloud in the grid. I use > > ``` > ` points = self.world2grid(points) > points.clamp_min_(0.0) > for...

Same problem with yours, It is really unpleasant!!

> 请问楼主解决了吗？ 未解决，后面手动改的

> Hello, thanks for the wonderful work! > > Sorry to bother, I'm trying to use bvh ray tracing to calculate the shadow effect, and I'm a little confused about...

> [@Dshijie](https://github.com/Dshijie) 此外，我们发现，直接在输入图像中拼接成一张（1 x N）的图可实现还不错的效果，但这并不是端到端训练过的推理方式，因此还有一点不稳定。一个优化思路是，用多图拼接的形式生成大批数据集，过滤出高质量的数据，用于训练。 Can we use this feature to extend the Flux-Kontext model to sequential model(video model)?