poly-diffuse

poly-diffuse copied to clipboard

【HD MAPPING】GT and input from init maptr seems not a same time

2

I have get the test result based on hd mapping branch with nuscenes mini dataset. But I found GT and input from init maptr seems not a same time reviewing...

yHsi

change supported image size in floorplan reconstruction task

1

Hello, I wan't to train the poly-diffuse with my own dataset, but my dataset is not suitable for scaling to 256*256, how can I change the poly-diffuse's code to support...

920902021

Generate density images from point cloud top view

1

Hello, Thanks for sharing the work. I want to run the inference pipeline and have images of the point cloud from the top view. How to create a density map...

aditdoshi333

File "train.py", line 191, in main() File "/root/anaconda3/envs/py3-mink/lib/python3.7/site-packages/click/core.py", line 1130, in __call__ return self.main(*args, **kwargs) File "/root/anaconda3/envs/py3-mink/lib/python3.7/site-packages/click/core.py", line 1055, in main rv = self.invoke(ctx) File "/root/anaconda3/envs/py3-mink/lib/python3.7/site-packages/click/core.py", line 1404, in invoke...

CBQ-1223

standard dm的一些实现的疑问和请教

2

 大佬您论文中说的standard dm的实现，有几个具体细节想请教您，因为想复线一下： 1.您是在room上做的没在maptr上实验哈 2.我理解你说的就是训练时 standard dm的实现：相当于改变两个数值，sigma_guide = 1.0,mu_guide = 0.0,然后想请教下损失函数是没有变化吗，还是每个gt加噪的实例预测X0然后和原本这个gt的坐标算loss，因为GS-DM没有训练分类损失， standard dm也没有这部分吗 3.推理时的standard dm，按论文中说的是从标准高斯噪声采样，那么这是坐标，请问类别信息会预测吗，具体是怎么得到类别预测的呢 超级感谢大佬答疑解惑

CBQ-1223

Why is the dataset only about ten gigabytes

2

I saw that the dataset in your github is 36GB, but after I compressed and decompressed it from the corresponding address, it was only 11GB. After putting it into the...

chaojun-gengboy

数据和模型有其他下载地址吗？

3

是否支持其他地址的下载？

lt573152873

MapTR预训练权重如何得到的

1

在训练denoise部分的时候，MapTR的forward的输入有timestep，在单独训练MapTR的时候，如何设置timestep的呢？

lvruichen

训练模型

2

我想利用自己的模型产生的平面图重建结果对polydiffuse进行训练，应该如何处理我产出的结果，数据集仍然是struct3d，或者你可以告诉我数据文件夹的文件的含义吗？万分感谢！

Monster-717