Open cjhyy7190 opened 5 years ago
it mentioned : pc_array = np.array([[x, y, z] for x,y,z in pc]) ValueError: too many values to unpack
excuse me , I adjust the pointcloud and the votenet can detect it.But the votenet always detect nothing.I used the nyu v2 depth dataset to aquire a pointcloud. the rgb image look like this.
@cjhyy7190 Is your pointcloud in upright-depth system(z is up,x is right,y is forward)?
@cjhyy7190 Is your pointcloud in upright-depth system(z is up,x is right,y is forward)?
maybe it is not.I am so sorry that I am not very familiar with the PCL library.And I also check out some information about the system.But I got nothing about this.Could you please tell me how to set the upright-depth system?I think it woule help me a lot. I want to have more communication with you.If you are convenient my e-mail is 335029901@qq.com.Thank you for your reply.
@cjhyy7190 Is your pointcloud in upright-depth system(z is up,x is right,y is forward)?
maybe it is not.I am so sorry that I am not very familiar with the PCL library.And I also check out some information about the system.But I got nothing about this.Could you please tell me how to set the upright-depth system?I think it woule help me a lot. I want to have more communication with you.If you are convenient my e-mail is 335029901@qq.com.Thank you for your reply.
@cjhyy7190 It seems that NYU does not provide camera extrinsics, so you need to manually adjust the pointcloud to the upright-depth system, such as using meshlab.
It seems that NYU does not provide camera extrinsics, so you need to manually adjust the pointcloud to the upright-depth system, such as using meshlab.
Thanks for your reply.And I am so sorry to trouble you again.I try to use meshlab to adjust the pointcloud,but i failed.I don't know how to use the meshlab to set the axis. Could you please tell me how to do this? Is that means adjust the axis by rotating the point cloud?Thank you very much!
I use a pointcloud without color like this.But the network still detect noting.But I think this thing is very clearly.I still cannot understand about the upright-depth system.How can I set the system?thanks for your reply.
HI @cjhyy7190
You have to set the axis direction right otherwise the prediction will be meaningless. In default the camera coordinate system has Y pointing downward, so it's very different from the required axis directions.
In Meshlab you can display the axis (see figure below) to have a check.
thank you very much.I have already solved this problem.But I also have some problems.sometimes the votenet can't detect the object accurately, the bounding box is so big or so small.sometimes there is nothing , but the votenet also detect something.Could you please tell me how to solve this problem?And when we acquired the bounding box , I find the color of the bounding boxes is always white.So how can we know the type of the object?Thank you for your reply.
Hello, Did you manually adjust the point cloud by using Meshlab or some other software? Could you give me some hints on changing the coordinates? Really appreciate for your help.
Hello, Did you manually adjust the point cloud by using Meshlab or some other software? Could you give me some hints on changing the coordinates? Really appreciate for your help.
you can use Meshlab to rotate your point cloud to adjust the coordinates
Hello, Did you manually adjust the point cloud by using Meshlab or some other software? Could you give me some hints on changing the coordinates? Really appreciate for your help.
you can use Meshlab to rotate your point cloud to adjust the coordinates
Thanks for your quick reply.
Hello,have you solved this problem?I meet similar problem.Do you know where should I modify when I want to detect only one class in SUNRBGD,such as box class?Wish your help,thx. 0.0
excuse me, I am very sorry to trouble you again.When I use the votenet to detect my own .ply file.It always mentioned me that question?How to modify this mistake?what should I do ? thank you for your reply.