Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

disparity pseudo-ground-truth #40

Open
Jianqiang-Mei opened this issue Mar 10, 2022 · 16 comments
Open

disparity pseudo-ground-truth #40

Jianqiang-Mei opened this issue Mar 10, 2022 · 16 comments

Comments

@Jianqiang-Mei
Copy link

请问作者大大论文中的disparity pseudo-ground-truth 那个图是怎么生成的呀,在哪里可以看到

@ootts
Copy link
Member

ootts commented Mar 12, 2022

如果你是指Fig.5第三行,用plt.imshow(mask*disp,'jet');plt.show()

@Jianqiang-Mei
Copy link
Author

这个我搞定了,想顺便问下大佬,怎么直接输出训练好的stereo_mask_rcnn预测的mask结果呀,我想做个对比试验,谢谢你

@Jianqiang-Mei
Copy link
Author

因为我发现vob和pob不同训练次数在val上的AP2d精度居然一样,所以想输出mask看一下

@ootts
Copy link
Member

ootts commented Mar 12, 2022

2d detector的box branch用的stereo rcnn。已经fix住了,没有训练。

@ootts
Copy link
Member

ootts commented Mar 12, 2022

这个参考visualiz.ipynb

@Jianqiang-Mei
Copy link
Author

我发现disprcnn对于很多非小轿车形状的车辆的预测,其实例视差都预测为小轿车的形状,请问这时受pseudo-ground-truth训练的影响吗

@Jianqiang-Mei
Copy link
Author

如果只用稀疏的lidar点云进行训练,是不是形状估计会更精确一点

@Jianqiang-Mei
Copy link
Author

考虑到实际场景中还会有卡车、货车这种和小轿车形状差异较大的车辆

@Jianqiang-Mei
Copy link
Author

2d detector的box branch用的stereo rcnn。已经fix住了,没有训练。

可是如果只训练了idispnet,这并不会影响mask的预测呀,请问你们是先用pseudo-ground-truth训练好了stereo_rcnn,然后再将其fix的吗

@Jianqiang-Mei
Copy link
Author

问题有点多,希望大佬能解惑一下,谢谢呀

@ootts
Copy link
Member

ootts commented Mar 12, 2022

  1. disprcnn对于很多非小轿车形状的车辆的预测,其实例视差都预测为小轿车的形状 是因为训练数据中多小轿车,卡车被kitti evaluation忽略,因此我们也没有训练卡车。
  2. 只训练了idispnet,这并不会影响mask的预测呀?不会。mask是在smrcnn里的mask branch.在idispnet之前。
  3. 稀疏的lidar使得监督信号sparse,尤其是比较远的车辆。不利于idispnet的训练

@Jianqiang-Mei
Copy link
Author

谢谢大佬

@Jianqiang-Mei
Copy link
Author

2d detector的box branch用的stereo rcnn。已经fix住了,没有训练。

既然stereo rcnn被fix住了,有什么方法可以在新的双目图像数据集上做可视化预测吗

@ootts
Copy link
Member

ootts commented Mar 13, 2022

inference是可以的

@Jianqiang-Mei
Copy link
Author

inference是可以的
请问能说下具体怎么操作吗

@Jianqiang-Mei
Copy link
Author

我试了,好像只能得到roi区域的inference结果,大佬能说下mask区域怎么inference的吗

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants