# NVIDIA-AI-IOT/Lidar_AI_Solution

A project demonstrating Lidar related AI solutions, including three GPU accelerated Lidar/camera DL networks (PointPillars, CenterPoint, BEVFusion) and the related libs (cuPCL, 3D SparseConvolution, YUV2RGB, cuOSD,).

Repository: https://github.com/NVIDIA-AI-IOT/Lidar_AI_Solution
Canonical: https://ross.abutalabs.com/products/lidar_ai_solution
Language: C++
License: NOASSERTION
License Family: other
Last push: 2026-07-28T09:51:55+00:00

## Health v2 (maintenance only)
Score: 72/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 94, release rhythm 35, longevity 86
- inputs: {"age_days": 1206, "days_push": 36, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1867, forks 338 (observed 2026-08-28T04:05:46.336348+00:00)

## What it is
NVIDIA's collection of GPU-accelerated Lidar AI inference solutions for autonomous driving, including optimized implementations of PointPillars, CenterPoint, and BEVFusion networks built on CUDA and TensorRT. It also ships supporting libraries such as cuPCL, 3D SparseConvolution, YUV2RGB, and cuOSD for point cloud processing and visualization.

## Use cases
- run lidar 3d object detection on gpu with pointpillars
- deploy bevfusion camera-lidar fusion inference with tensorrt
- accelerate centerpoint inference with cuda and int8 quantization
- export 3d sparse convolution backbone from onnx for fast inference
- voxelize lidar point clouds with cuda kernels
- convert yuv camera frames to rgb on gpu for sensor fusion pipelines
- draw bounding boxes on lidar point cloud visualizations with cuosd
- quantize mmdet3d spconv models for edge deployment

## When to choose
- you need production-grade, GPU-accelerated inference for lidar-based 3D detection networks like PointPillars, CenterPoint, or BEVFusion
- you are deploying perception pipelines on NVIDIA GPUs or Jetson edge devices and want TensorRT/CUDA optimization
- you need low-latency voxelization, NMS, or preprocessing kernels implemented in CUDA
- you want quantization (PTQ/QAT) workflows for sparse convolution lidar backbones

## When to avoid
- you need a general-purpose deep learning framework rather than lidar-specific inference solutions
- your hardware is not NVIDIA CUDA-capable
- you want training-first tooling with minimal inference focus (though some training support exists, the project centers on deployment)
- you need a plug-and-play solution without C++/CUDA integration effort

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, llm-inference, computer-vision, image-processing, gpu-computing, sdk, developer-tools
- domain: autonomous-vehicles, deep-learning, computer-vision, gpu-computing, robotics, artificial-intelligence
- platform: cpp
- tags: lidar, point-clouds, 3d-object-detection, bevfusion, centerpoint, pointpillars, sparse-convolution, tensorrt, onnx, cuda-kernels, self-driving, sensor-fusion, int8-quantization, fp16, linux, cuda, gpu, nvidia-jetson

## Member repositories
- NVIDIA-AI-IOT/Lidar_AI_Solution (main) score 72

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:46.336348+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:15:22.667968+00:00, confidence not recorded.
  - readme: https://github.com/NVIDIA-AI-IOT/Lidar_AI_Solution (fetched 2026-08-28T04:05:46.336348+00:00, sha 58e6453c5c5b)
- Data as of 2026-08-30T08:39:29.467469+00:00.