PointNet
Last updated
Was this helpful?
Last updated
Was this helpful?
ํ์ฉ ๋ฐ์ดํฐ๋ท
ํ์ต : Semantic3D
์ถ๋ก : Semantic3D + KITTI
During both training and inference, PointNet++ is fed with fix-sized cropped point clouds within boxes, we set the box size to be 60m x 20m x Inf, with the Z-axis allowing all values. During inference with KITTI, we set the region of interest to be 30m in front and behind the car, 10m to the left and right of the car center to fit the box size. This allows the PointNet++ model to only predict one sample per frame.
์ฐจ๋ณ์ : In PointNet++โs set abstraction layer, the original points are subsampled, and features of the subsampled points must be propagated to all of the original points by interpolation
๊ธฐ์กด : This is achieved by 3-nearest neighbors search (called ThreeNN)---
๋ณ๊ฒฝ : Open3D uses FLANN to build KDTrees for fast retrieval of nearest neighbors, which can be used to accelerate the ThreeNN op.
Inference on sparse pointcloud (KITTI)
Inference results after interpolation
The sparse labels need to be interpolated to generate labels for all input points. This interpolation can be achieved with nearest neighbor search using open3d.KDTreeFlann and majority voting, similar to what we did above in the ThreeNN op.
๋ณด๊ฐ(interpolation)์์ ์ ์ ์ฒด ์์ ์๊ฐ์ 90%๋ฅผ ์ฐจ์งํ๊ณ , 1FPS์ ์๋๋ฅผ ๋ณด์ธ๋ค. ํด๊ฒฐ์ ์ํด custom TensorFlow C++ op InterploateLabel๋ฅผ ์ ์ฉํ์ฌ 10+FPS์๋ ํฅ์์ ๋ณด์๋ค.
๊ถ์ฅ ํ๊ฒฝ : ubuntu 16.04, cuda 9.0, python3, tensorflow 1.2 (1.8๊ถ์ฅ), open3d 0.6+
build TF ops. Youโll need CUDA and CMake 3.8+.
[์ฝ๋] Semantic3D semantic segmentation with Open3D and PointNet++, [๊ฒฐ๊ณผ] : [์ฐธ๊ณ ์ฝ๋]๋ฅผ ํ์ฉํ์ฌ Open3D๋ก ์์ฑ๋ ์ฝ๋
[์ฐธ๊ณ ์ฝ๋] PointNet2 for semantic segmentation of 3d points clouds, [๊ฒฐ๊ณผ] : [์๋ณธ์ฝ๋]๋ฅผ ํ์ฉํ์ฌ์ ์์ฑ๋ ์ฝ๋