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

关于Reszie3D #5

Open
betterhalfwzm opened this issue Aug 24, 2022 · 7 comments
Open

关于Reszie3D #5

betterhalfwzm opened this issue Aug 24, 2022 · 7 comments

Comments

@betterhalfwzm
Copy link

您好,训练的时候用的dict(type='Resize3D', img_scale=(640, 384), keep_ratio=True),,测试的时候应该怎么用Reszie3D啊?

@Tai-Wang
Copy link
Owner

在 RandomCrop 之前加同样的 resize 即可。

@betterhalfwzm
Copy link
Author

betterhalfwzm commented Aug 24, 2022

在 RandomCrop 之前加同样的 resize 即可。
@Tai-Wang
训练的pgd,跑mono_det_demo.py时,3D框位置不对
image
2cac95007c22ed765c25340f1549262
f6c35b8d36e507c87b9450e7989e307
训练ap:
----------- AP11 Results ------------

Pedestrian [email protected], 0.50, 0.50:
bbox AP11:16.3295, 8.1563, 8.1664
bev AP11:9.0909, 9.0909, 9.0909
3d AP11:9.0909, 9.0909, 9.0909
aos AP11:14.75, 7.39, 7.39
Pedestrian [email protected], 0.25, 0.25:
bbox AP11:16.3295, 8.1563, 8.1664
bev AP11:17.9982, 9.0909, 9.0909
3d AP11:17.9982, 9.0909, 9.0909
aos AP11:14.75, 7.39, 7.39
Cyclist [email protected], 0.50, 0.50:
bbox AP11:9.0909, 9.0909, 9.0909
bev AP11:4.5455, 9.0909, 9.0909
3d AP11:0.0000, 9.0909, 9.0909
aos AP11:9.05, 9.09, 9.09
Cyclist [email protected], 0.25, 0.25:
bbox AP11:9.0909, 9.0909, 9.0909
bev AP11:9.0909, 9.0909, 9.0909
3d AP11:9.0909, 9.0909, 9.0909
aos AP11:9.05, 9.09, 9.09
Car [email protected], 0.70, 0.70:
bbox AP11:18.0936, 9.0909, 9.0909
bev AP11:17.2969, 9.0909, 9.0909
3d AP11:16.8397, 9.0909, 9.0909
aos AP11:18.09, 9.09, 9.09
Car [email protected], 0.50, 0.50:
bbox AP11:18.0936, 9.0909, 9.0909
bev AP11:17.7301, 9.0909, 9.0909
3d AP11:17.7033, 9.0909, 9.0909
aos AP11:18.09, 9.09, 9.09

Overall AP11@easy, moderate, hard:
bbox AP11:14.5047, 8.7794, 8.7827
bev AP11:10.3111, 9.0909, 9.0909
3d AP11:8.6435, 9.0909, 9.0909
aos AP11:13.96, 8.52, 8.52

----------- AP40 Results ------------

Pedestrian [email protected], 0.50, 0.50:
bbox AP40:8.9505, 6.7290, 4.4888
bev AP40:6.2250, 4.2130, 4.2130
3d AP40:5.9250, 4.0278, 4.0278
aos AP40:8.11, 6.10, 4.06
Pedestrian [email protected], 0.25, 0.25:
bbox AP40:8.9505, 6.7290, 4.4888
bev AP40:9.7980, 7.3611, 7.3611
3d AP40:9.7980, 7.3611, 7.3611
aos AP40:8.11, 6.10, 4.06
Cyclist [email protected], 0.50, 0.50:
bbox AP40:0.0000, 2.5000, 2.5000
bev AP40:0.0000, 1.6667, 1.6667
3d AP40:0.0000, 0.0000, 0.0000
aos AP40:0.00, 1.68, 1.68
Cyclist [email protected], 0.25, 0.25:
bbox AP40:0.0000, 2.5000, 2.5000
bev AP40:0.0000, 2.5000, 2.5000
3d AP40:0.0000, 2.5000, 2.5000
aos AP40:0.00, 1.68, 1.68
Car [email protected], 0.70, 0.70:
bbox AP40:14.8795, 7.4528, 7.4361
bev AP40:13.7993, 7.0012, 6.9478
3d AP40:13.1073, 6.6759, 6.6305
aos AP40:14.87, 7.44, 7.40
Car [email protected], 0.50, 0.50:
bbox AP40:14.8795, 7.4528, 7.4361
bev AP40:14.4030, 7.2744, 7.2469
3d AP40:14.3846, 7.2704, 7.2389
aos AP40:14.87, 7.44, 7.40

Overall AP40@easy, moderate, hard:
bbox AP40:7.9433, 5.5606, 4.8083
bev AP40:6.6748, 4.2936, 4.2758
3d AP40:6.3441, 3.5679, 3.5527
aos AP40:7.66, 5.07, 4.38

2022-08-24 11:28:56,915 - mmdet - INFO - Results of img_bbox2d:

----------- AP11 Results ------------

Pedestrian [email protected], 0.50, 0.50:
bbox AP11:0.0000, 0.0000, 0.0000
Pedestrian [email protected], 0.25, 0.25:
bbox AP11:0.0000, 0.0000, 0.0000
Cyclist [email protected], 0.50, 0.50:
bbox AP11:0.0000, 0.0000, 0.0000
Cyclist [email protected], 0.25, 0.25:
bbox AP11:0.0000, 0.0000, 0.0000
Car [email protected], 0.70, 0.70:
bbox AP11:0.0000, 0.0000, 0.0000
Car [email protected], 0.50, 0.50:
bbox AP11:0.0000, 0.0000, 0.0000

Overall AP11@easy, moderate, hard:
bbox AP11:0.0000, 0.0000, 0.0000

----------- AP40 Results ------------

Pedestrian [email protected], 0.50, 0.50:
bbox AP40:0.0000, 0.0000, 0.0000
Pedestrian [email protected], 0.25, 0.25:
bbox AP40:0.0000, 0.0000, 0.0000
Cyclist [email protected], 0.50, 0.50:
bbox AP40:0.0000, 0.0000, 0.0000
Cyclist [email protected], 0.25, 0.25:
bbox AP40:0.0000, 0.0000, 0.0000
Car [email protected], 0.70, 0.70:
bbox AP40:0.0000, 0.0000, 0.0000
Car [email protected], 0.50, 0.50:
bbox AP40:0.0000, 0.0000, 0.0000

Overall AP40@easy, moderate, hard:
bbox AP40:0.0000, 0.0000, 0.0000

@Tai-Wang
Copy link
Owner

请问您用的哪个config?看起来不像是repo里给定的。FCOS3D或者PGD在KITTI上的Resize3D如何使用还有待测试,如有需要可以先用DfM和Waymo上PGD的config尝试。

@betterhalfwzm
Copy link
Author

请问您用的哪个config?看起来不像是repo里给定的。FCOS3D或者PGD在KITTI上的Resize3D如何使用还有待测试,如有需要可以先用DfM和Waymo上PGD的config尝试。
@Tai-Wang
pgd kitti那个config,换了个backbone,其他的应该一样
我尝试了训练原图,测试640384,3d位置不对;
image
训练图片640
384,测试640384,3d位置不对;
image
训练图片640
384,测试原图,3d位置不对;
image
只有原图训练原图测试3d位置才对的上
image

@betterhalfwzm
Copy link
Author

betterhalfwzm commented Aug 24, 2022

当scale_factor=0.515时,是这个效果,训练图片640384,测试640384,kitti的原图是1242, 375,必须保持同比例缩放吗?
9f8e11949dd5887decf327153706185
image

@betterhalfwzm
Copy link
Author

betterhalfwzm commented Aug 26, 2022

pgd 训练nus,@Tai-Wang
image
train:
image

test:
image

@Hukongtao
Copy link

@betterhalfwzm 你好,关于Resize3D在测试的时候怎么用的问题您这边解决了么?

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

3 participants