doubleZ

Hi, this problem also occurs when I use the NO `inverse_depth` flag for training. After digging into the error message, I find that the original code provided by this repo...

> Original size `1600x1184` cause OOM on my 11GB GPU. I resized the image to `1152x864` and it works (costs 6831MB). Don't forget to change the instrinsics as the following:...

Kindly link to => https://github.com/colmap/colmap/issues/570#issuecomment-1960852296

Hi, there. I also meet the problem when offscreen render with texture, here is my code ```python tex = pyrender.Texture(source=Image.open('mytexture.png'), source_channels='RGB', width=width, height=height) mat = pyrender.material.MetallicRoughnessMaterial(normalTexture=tex, wireframe=False) ref_scan = load_obj(obj_path)...

@submagr Sorry for the late reply. I have send the code & model to your personal google email. Thanks for your work!

I have also meet this error when testing on the DTU dataset. ``` err, "a Variable name or other graph key that is missing") tensorflow.python.framework.errors_impl.NotFoundError: Restoring from checkpoint failed. This...

Finally, I train it myself and everything just go on the right way. Maybe there are something changed in the pretrained ckpt file from the author side?

@chinedu0507 he said just right, you can comment out `find_package(Boost REQUIRED python37)` and use `find_package(Boost REQUIRED python)` below instead. Then in my case I fix my error. Hope this state...

这个问题问法有点奇怪，mvs网络依次将每张图片作为参考图像输入网络，得到每张图像对应的多视角深度图，最终再将深度图融合成点云，如果你问的是最终点云的坐标系是以哪里作为原点参考，这个取决于colmap提取出的相机pose是以谁为原点的

@Artiprocher 我也遇到了类似的问题，我使用Wan2.1-VACE-14B和Wan2.1-VACE-1.3B进行测试 单卡推理正常结果 https://github.com/user-attachments/assets/5009ebdd-60c0-4489-a7c4-c639507dfc7d 但多卡开usp推理后半段会有噪声 https://github.com/user-attachments/assets/aa665de3-3759-478b-984b-cb59dce9af7d 目前问题定位到usp的问题 补充：之前测过Wan-T2V的usp没问题，不知道使用其他模型开usp时问题在哪里，期待解答