Train 3D detectors with LiDAR and camera fusion.
What it does
Trains, evaluates, and runs inference with BEVFusion models for 3D object detection using LiDAR and camera data.
When to use it
Use when training, evaluating, or running inference with a TAO BEVFusion model.
How to use it
Give it a BEVFusion training, evaluation, or inference request and dataset files; it runs the selected action and produces results files.
What you provide
- Your files
Uses
What you need · 7
Requires Docker to run the BEVFusion container.
Requires NVIDIA Container Toolkit alongside Docker.
Requires the BEVFusion-specific TAO 5.5.0 container; shared TAO PyTorch 7.x images lack mmdet3d.
Specifies 24GB+ VRAM per GPU, with A100 recommended. One GPU supports small smoke or AutoML validation runs; ordinary training should use two or more.
Access · 0
No additional actions listed.