Cloudini: pointcloud compression for the rest of us

[facontidavide]

Hi,

I am please to announce the first decent release of Cloudini, the pointcloud compression library that you missed in your life.

The focus of this library is FAST compression and decompression, where the primary applications are:

• decrease (sometimes dramatically) the dimension of your rosbags containing sensor_msgs/msg/PointCloud2
• allow faster streaming over the network.
• pay less the cloud provider that charge for the Gb…

What to expect

1. It is a lossy algorithm, but don’t you worry, you can customize the resolution. 1 mm? You will be good. Felling paranoid? Use 0.1 mm and you still get very good compression ratio.

2. Compression ratio in the dataset I used have a range from 0.15 to 0.5 of the original size (smaller is better).

3. Faster and better compression ratios than using LZ4 or ZSTD alone.

4. Yes, I did try Draco , and in my limited benchmarks, Cloudini has similar compression ratio but is considerably faster.

5. Preliminary implementation of point_cloud_transport plugins… but it needs more testing

6. Simple and small. Enjoy reading the code and learning how simple the algorithm is; I did try more complex ones from the literature, but the simplest performed better.

Do you want to test it RIGHT NOW?

Download the rosbag encode here and try encoding your own rosbag (MCAP required).

You may need to run:

chmod +x cloudini_rosbag_converter-*.AppImage

once to execute the file.

Then run:
./cloudini_rosbag_converter-v0.1.0-x86_64.AppImage -c -f your_own_rosbag.mcap -o encoded_bag.mcap

[facontidavide]

To have some fun, and learn some WebAssembly, I created a simple web app that will run Cloudini in your browser and report the potential compression ratio.

Note that if your rosbag is already using ZSTD compression, the saving will NOT be that dramatic.

Try it yourself here: https://cloudini.netlify.app/