Browse categories
Explore
Fiverr Pro
English
$
USD
I will build 3d lidar point cloud object detection models like pointpillars
Daksh A
About this gig
Need 3D perception for autonomous vehicles, robots, drones, or smart infrastructure? I specialize in LiDAR point cloud object detection, a skill few freelancers offer.
Models I work with:
- PointPillars, SECOND, and VoxelNet (voxel and pillar based)
- CenterPoint (center-based 3D detection and tracking)
- BEVFusion (LiDAR and camera fusion in bird's-eye view)
- PointNet++ and PV-RCNN
What I do:
- Train or fine-tune 3D detectors on KITTI, nuScenes, Waymo-format, or your own custom LiDAR data
- Convert custom datasets (PCD, BIN, LAS, ROS bags) into training-ready formats
- Tune anchors, voxel sizes, and point cloud ranges for your sensor
- Evaluate with 3D/BEV mAP and NDS, with per-class analysis
- Visualize 3D boxes in Open3D or RViz
- Optimize for real-time inference with TensorRT and ROS/ROS2 nodes
Every order includes:
- A data review and model recommendation
- Trained checkpoints and configs
- An evaluation report with visual results
- A documented, reproducible training and inference setup
Stack: Python, PyTorch, OpenPCDet, MMDetection3D, Open3D, CUDA, ROS2.
Please message me with your sensor type and data format before ordering.
Get to know Daksh A
Daksh A
Machine Learning Engineer
- FromIndia
- Member sinceDec 2020
- Avg. response time1 hour
Languages
English
I am a Machine Learning Engineer with extensive experience in building production-grade AI pipelines, predictive failure detection models, and custom RAG systems. I specialize in MLOps, computer vision, and LLM evaluation, having successfully deployed models that significantly lift accuracy and operational efficiency across diverse enterprise environments.
Other AI Development Services I Offer
FAQ
Do you provide GPU compute?
Yes for small runs. Large training jobs run on your cloud or GPU machine, and I'll estimate the hours upfront.
My data isn't labeled. Can you help?
I can advise on 3D labeling tools and formats, or quote 3D box annotation separately. Pretrained models can also pre-label your data.
Which LiDAR sensors are supported?
Any sensor that exports point clouds (Velodyne, Ouster, Livox, Hesai, and others). I adapt the point range, voxel size, and intensity handling to your sensor.
Can you add camera fusion?
Yes. The Premium package covers BEVFusion-style LiDAR and camera fusion if you have calibrated camera data.
