How To Visualize LiDAR Point Cloud In Python | KITTI Dataset

[@LightsCameraVision]

If you know a better repository/package/software for such visualization, please do share here. Thanks for watching. ✌️

[@Naveen-jr2io]

Thanks a lot sir. It was a really useful video to visualize point cloud from a public dataset. It will be helpful if you could get a video regarding point cloud segmentation processing

[@LightsCameraVision]

Thank you for the kind words. I have plans to make segmentation videos. ✌️

[@myproject7762]

How to do custom object detection on point cloud data ??

[@LightsCameraVision]

1. PointNet is a well-known architecture for 3D object detection. I covered the PointNet concept and explained the code in this video: kzbin.info/www/bejne/fnq1m2imfZenY7s Although I covered it for the classification task, it is also known for segmentation and detection. The detector and classifier share the same architecture, almost. I shared the PointNet paper and code link in the video description of the above link. You can read this GitHub issue for how to prepare your custom data: github.com/charlesq34/pointnet/issues/75 2. Another popular architecture for point cloud object detection is VoxelNet. Paper: arxiv.org/pdf/1711.06396.pdf Code: github.com/qianguih/voxelnet I intend to make a point cloud object detection video in the near future. I hope this helps. Feel free to comment if you have any other questions. Happy Learning!

[@yasminbanu2597]

Hello sir what I have searched in internet I found the same in ur video. Thanks for that. It will be helpful if you make video regarding how to give point cloud segmented images to yolo network

[@LightsCameraVision]

I’m very glad that you found this video helpful. 🙂 Thanks for watching and for the suggestion. A point cloud segmentation video is in the making. ✌️

[@yasminbanu2597]

Thank you for your reply

[@sumedhvidhate3383]

conda: command not found getting this error

[@LightsCameraVision]

So this may mean many things. You may not have anaconda installed or maybe conda isn't added in the path variable. If you don't have anaconda installed go to www.anaconda.com/ and follow the instructions to download it. If you already have it downloaded you have to follow something like this monovm.com/blog/conda-command-not-found-fixed/ or other StackOverflow posts that fits your case. Then follow the video to create the environment and install the necessary dependencies. ✌️

[@sumedhvidhate3383]

I installed conda and after installing mayaavi package I was still getting other error called as opengl error

[@danielmathew6961]

Thanks for the video, it is very helpful and I really appreciate it! Is there a way to download just a subset of the KITTI dataset so I don't have to download the entire dataset?

[@LightsCameraVision]

I'm glad that you found it helpful. Here is a small subset of the KITTI dataset that you can download. pl-flash-data.s3.amazonaws.com/KITTI_tiny.zip ✌️

[@danielmathew6961]

@@LightsCameraVision Thanks so much for the dataset! Is it also possible to get the images for the subset you gave me? The script doesn't run without the images corresponding with the labels, calibration, and pointcloud data. If not, is there another way to run the script without the images?

[@LightsCameraVision]

Hi Daniel, try this subset of KITTI data. drive.google.com/drive/folders/1Y6QKWAEN0mUuz2lq7D1pJNCsg2cnwhjC?usp=sharing

[@sumedhvidhate3383]

conda install -c anaconda -c conda-forge mayavi is the updated package for installing mayaavi it will solve all the errors

[@LightsCameraVision]

I would suggest you to fix your conda not found error first and then install the packages/dependencies. You can run "conda --version" to check if it's installed properly.

[@sumedhvidhate3383]

Install it 3 to 4 times

[@sumedhvidhate3383]

And check to still got same error

[@houdaa1810]

hello, thanks for this vedio, i want to know if we can use the kitti dataset with pointnet model ??

[@LightsCameraVision]

Appreciate it. You can use it. You may need to change the code a little bit. Please check this repo github.com/kargarisaac/PointNet-SemSeg-VKITTI3D They used PointNet for semantic segmentation on Virtual KITTI dataset. It should give you some idea. ✌️

[@yasminbanu2597]

Sir i tried the same but while running the show command my output window gets disappear within a second. I can't able to see anything in window

[@LightsCameraVision]

Are you getting any error?

[@yasminbanu2597]

@@LightsCameraVision No error sir

[@LightsCameraVision]

​@@yasminbanu2597 I'm assuming you are using the same GitHub repository for KITTI data. Then make sure you have installed all the necessary libraries and are using the correct command (like below) in the terminal/shell. python kitti_object.py --show_lidar_with_depth --vis