Comments (7)
Thanks for reporting this error. I think the position is correct. I generate the index in the 2D matrix in this way:
centerformer/det3d/models/necks/rpn_transformer.py
Lines 247 to 250 in 5a949b8
Here
neighbor_coords[:, :, :, 1]
comes from the y_coord
and neighbor_coords[:, :, :, 0]
comes from the x_coord
.
However, the bug is probably this:
centerformer/det3d/models/necks/rpn_transformer.py
Lines 244 to 246 in 5a949b8
The coordinates should be clamped separately using their own size rather than just assuming
H = W
.
I don't have a machine at hand to test. Can you check if changing this can fix your error?
from centerformer.
Hello,
I notice that, the bug reporting can disappear through x_coor and y_coor being clamped within [0, H] and [0, W]. But it could probably leave the fact behind that x_coor = order // W and y_coor = order % W
. I did a small test:
According to the picture, the revised version seems correctly make x_coor = 7 and y_coor =4.
Could you think this over and recheck the implementation when you are available? Of course, if "transposing" or some other purposes exist, I would apologize for my misunderstanding and bothering you.
from centerformer.
Oh, I see. So the x_coor
and y_coor
here do not mean the row and col indexes of this value in the 2D matrix (like [7,4]). Conversely, it means the indexes on the height dimension (y_coor
) and width dimension (x_coor
), which are common to describe the coordinates of a pixel in an image.
I will recheck the code just in case there are some other bugs.
from centerformer.
Oh, I see. So the
x_coor
andy_coor
here do not mean the row and col indexes of this value in the 2D matrix (like [7,4]). Conversely, it means the indexes on the height dimension (y_coor
) and width dimension (x_coor
), which are common to describe the coordinates of a pixel in an image.I will recheck the code just in case there are some other bugs.
Since height dimension (y_coor
) and width dimension (x_coor
), in Line 469, it should be torch.stack([y_coor, x_coor], dim=2)
, right?
from centerformer.
The order of the x_coor
and y_coor
variables does not matter in this case, as long as you remember the correct dimension that each variable represents. The deformable attention layer also requires storing the reference point position in the same way.
from centerformer.
Oh, I see. I will go through it again and recheck it. Thank you so much for the clarification.
from centerformer.
Great! If you have any further questions or concerns, feel free to reopen the issue.
from centerformer.
Related Issues (20)
- About `disable_dbsampler` HOT 2
- About Lidar and image fusion
- The effect of deformable attention HOT 1
- Question about why the add&norm structure of the tranformer network differ from the typical transformer one HOT 3
- Use CenterFormer on other datasets HOT 2
- CenterFormer on kitti
- Evaluation on waymo opendataset HOT 8
- waymo coordinates HOT 2
- previous frame transformed to current frame? HOT 1
- Positional embedding in RPN_transformer_deformable_multitask
- Is it correct that Nan appears in the loss? HOT 5
- AUTOMATIC MIXED PRECISION
- /usr/include/stdio.h(189): error: attribute "__malloc__" does not take arguments
- Redundant boxes after post processing
- Implementation of CorssAttention
- Are the x, y, z values included in the value of the variable center_pos?
- nuScenes result?
- Issue in points np.concatenate(s_points_list, axis=0) in centerformer-master/det3d/core/sampler/sample_ops.py HOT 2
- issues about testing on nuscenes testset
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from centerformer.