weiyithu/FGR

[ICRA 2021] FGR: Frustum-Aware Geometric Reasoning for Weakly Supervised 3D Vehicle Detection

40
/ 100
Emerging

This project helps self-driving car engineers and researchers automatically generate 3D bounding box labels for vehicles in lidar and camera datasets. It takes raw KITTI dataset sensor data (camera images, lidar point clouds, and calibration) and outputs estimated 3D bounding box labels for cars, which are then used to train 3D object detection models. This is useful for anyone working on autonomous driving perception systems who needs to efficiently create training data without extensive manual labeling.

No commits in the last 6 months.

Use this if you need to create training labels for 3D vehicle detection models from raw KITTI sensor data, particularly when ground truth labels are scarce or expensive to obtain.

Not ideal if you need to detect objects other than vehicles, or if your dataset is not in the KITTI format.

autonomous-driving 3D-object-detection lidar-data sensor-fusion training-data-generation
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 8 / 25
Maturity 16 / 25
Community 16 / 25

How are scores calculated?

Stars

44

Forks

8

Language

Python

License

MIT

Last pushed

May 18, 2021

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/computer-vision/weiyithu/FGR"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.