fabiopoiesi / dip Goto Github PK
View Code? Open in Web Editor NEWProject page of the paper "Distinctive 3D local deep descriptors" accepted in IEEE International Conference on Pattern Recognition 2020.
Project page of the paper "Distinctive 3D local deep descriptors" accepted in IEEE International Conference on Pattern Recognition 2020.
Hi!
When I run "train.py" using the preprocessed data, I got the error"OSError:Unable to open object (bad object header version number)". I don't know how to solve.Do you have any ideas please?Thank you very much.
Hi, I really appreciate your work on point cloud registration .The link of preprocessed data has been unable to be opened. Is there any other way or can you send it to me.Thank you very much.
Line 27 in 6b72d5c
ptnn
is arranged like [3, N], so it should be 1 / ptnn.shape[1]
instead of 1 / len(ptnn)
?Hi @fabiopoiesi ,
I really enjoyed reading your paper and thanks for supplying your code.
After reading the paper I have some question which are still unclear to me.
In the bottleneck part you explain how you find potential matched points within the corresponding patches using alpha
and alpha'
that their corresponding feature cross some threshold.
It is not clear to me when you use this idea in the training pipeline, or you just mentioned that for general analysis?
In Equ (6), it is not clear to me how you define the sets C_pos
and C_neg
and how you generate the feature vectors f
.
Does C_pos are a set of matching patches with their network output?
Thanks
Yuval
In my experiments, the loss remains to be around 0.8-1.1 even after 300k iterations. Is it right?
The ply of 3DMatch data from fepth-fusion.
When will App's apk be available?
Thanks for your sharing, here I have a question about how to generate 3DMatch_train.zip data.
As far as I know, the data downloaded from 3DMatch contains 62 RGB-D sequences, do those *.ply files were generated by TSDF fusion provided by 3DMatch-toolbox?
Thanks for your help.
Hi, thanks for your interesting work.
When I preprocessed 3DMatch training data, I got the error "AttributeError: module 'open3d' has no attribute 'registration' ".
from isl-org/Open3D#2951
**o3d.pipelines.**registration.compute_fpfh_feature(pcd).
It seems that now it is under pipelines based on the new documentation http://www.open3d.org/docs/release/python_api/open3d.pipelines.html
So in preprocess_3dmatch_correspondences_train.py, we can use:
result =o3d.pipelines.registration.registration_icp(pcd1, pcd2, .02, np.eye(4),
o3d.pipelines.registration.TransformationEstimationPointToPoint())
Thanks!
Best wishes!
Hello professor, can you provide the weights trained on KITTI? thank you very much!
Hi, I am really appreciated with your works!
I am just wondering how you dealt with the situation where no points or fewer than 3 points are inside the LRF kernel in lpf.py:
Line 21 in 6b72d5c
This results in complex numbers in xp, yp, zp.
Line 45 in 6b72d5c
How did you resolve this problem?
As a solution, putting in zero xyz values to the patch would be a problem. Any ideas?
Thank you very much!
Hi,I used the trained model ckpt_e39_i14400_dim32.pth and the final_chkpt.pth you gave for testing.The test results are different.May I ask how the final_chkpt.pth is obtained.
Hi!
When I run "preprocess_3dmatch_lrf_train.py", I got the error "AttributeError: 'open3d.cpu.pybind.geometry.PointCloud' object has no attribute 'select_down_sample'".
Using
# select only corresponding points
pcd1_corr = pcd1.select_by_index(corrs[:, 0])
pcd2_corr = pcd2.select_by_index(corrs[:, 1])
will fix the error.
Thanks.
Best wishes!
Hi,I'm very interested in your work.
How did you get the final_chkpt.pth?
I am looking forward to your reply.Thank you very much.
Hi, very appreciate the work you have done!
I am just curious about the license of this work. MIT? BSD?
Any plan to announce how you are going to distribute this work?
Thank you so much
Dear professor,
When I run the `Demo_eth.py` ,I get a bug.
E:\project_1\dip-master\lrf.py:47: RuntimeWarning: divide by zero encountered in double_scalars a1 = 1 / np.linalg.norm(np.dot(v, (alpha * beta)[:, np.newaxis])) E:\project_1\dip-master\lrf.py:49: RuntimeWarning: divide by zero encountered in double_scalars xp = 1 / np.linalg.norm(np.dot(v, (alpha * beta)[:, np.newaxis])) * np.dot(v, (alpha * beta)[:, np.newaxis]) E:\project_1\dip-master\lrf.py:49: RuntimeWarning: invalid value encountered in multiply xp = 1 / np.linalg.norm(np.dot(v, (alpha * beta)[:, np.newaxis])) * np.dot(v, (alpha * beta)[:, np.newaxis])
Can you give mo some advice? Thanks.
Best.
the result we got using 'python demo_3dmatch.py' is not so good as presented in the paper.The cups on tabel is not aligned.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.