izhengfan / se2clam Goto Github PK

thank you to release this program ,it is very impressive and useful .
but I have some problem with my dataset ,i tried to record image and odom data by ROS ,and I think the odom data means it have a origin point and record the relative relationship between the robot and the origin,so I get nav_msgs/Odometry->pose to use my odom data,but when i refer to your data,i realise that my profile is very different from yours,so i'm not sure whats wrong with my set,con you tell me your's data unit？

this is my dataset
https://drive.google.com/file/d/1WvbhH-GzAaZA3E2IITXCSom7MtctMbkz/view?usp=sharing

when i go stright robot almost not move,just turn looks correct.

thank you.

Why did I follow the steps in your demo to run it.
The result is different from the picture in your demo
No map appears in my rviz。

Nice writing! I want to get ground truth indoor as yours. Do you have more relevant papers on the method for obtaining the ground value using QR code?

Hi,
I'm reading the paper and try to understand the equations, but I have few questions about them. Could you please help me to figure it out if you are convenient?
1. In equation (5), the information matrix is denoted as lowercase omega, however, equation (9) uses both lowercase and capital omega. Do they have the same meaning?
2. Equation (23), the error in frame {B} is defined as a subtraction of two se3. Why not define it as the same form of equation (22) in camera frame?
3. Equation (46) is the calculated Hessian matrix "H" for the two keyframes after marginalization. According to my understanding, using H=J^(T) * Omega * J, the information matrix Omega could be calculated once H and J are known. In the paper, the equation (47) is used to calculate the information matrix with J+. However, I can't understand the construction of J+. Could you please make it more detailed?

I would appreciate if you could help me with the questions. Looking forward to you replying.

Your work is great ! Could this model be used on KITTI odometry dataset ? And how to make it ?

I have read part of source code, It seems no exists initialization process, which is different from ORB_SLAM2. and R t are directly get from odom-2d in triangulate. Local BA part is not read, and it seems use slide window size from what I see in rviz.right? Global BA is alsome and I am curious about it, can you explain it process theory/chart. sorry to bouther you.

Hi, I wanna run se2clam with my own dataset, and I have modified these files and uploaded them to the link:

• folder: image

• file: "odo_raw.txt"

• config files: "ComConfig.yml", "Settings.yml"

But when I run se2clam, trajectory cannot be showed correctly.

The figure below shows the wrong trajectory using my own dataset.

The real trajectory looks like this:

I would appreciate it if you could show light on these questions:

• whether I have modified files wrong or other reasons.

• the unit of the x and y in saved trajectory file "se2clam_kf_trajectory.txt".

const int levelCols = sqrt((float)nDesiredFeatures/(5imageRatio));
const int levelRows = imageRatiolevelCols;

above code pasted from ComputeKeyPoints() function. I want to kown why calc the two value in this way but they are const variable 30 in ORB_SLAM2 code.
hope for your reply. thanks

does tvec_b_c means camera to odom/vehicle frame transform? and its value in config file is
data: [123.9100, 399.1023, 0]
I want to ask what's the unit of this data, meter? seems too big.

郑帆您好：
早先在Youtube上看到过您讲解这个案例的视频，现在现在终于看到源码和论文。感谢感谢。我有个问题：在这类稀疏点云地图上如何做机器人的路径规划呢？我没找到比较合理的方案，希望您这里能给一些建议，谢谢啦：）