FCAF3D: Fully Convolutional Anchor-Free 3D Object Detection
Abstract
Recently, promising applications in robotics and augmented reality have attracted considerable attention to 3D object detection from point clouds. In this paper, we present FCAF3D --- a first-in-class fully convolutional anchor-free indoor 3D object detection method. It is a simple yet effective method that uses a voxel representation of a point cloud and processes voxels with sparse convolutions. FCAF3D can handle large-scale scenes with minimal runtime through a single fully convolutional feed-forward pass. Existing 3D object detection methods make prior assumptions on the geometry of objects, and we argue that it limits their generalization ability. To eliminate prior assumptions, we propose a novel parametrization of oriented bounding boxes that allows obtaining better results in a purely data-driven way. The proposed method achieves state-of-the-art 3D object detection results in terms of mAP@0.5 on ScanNet V2 (+4.5), SUN RGB-D (+3.5), and S3DIS (+20.5) datasets.
Introduction
We implement FCAF3D and provide the result and checkpoints on the ScanNet and SUN RGB-D dataset.
Results and models
ScanNet
| Backbone | Mem (GB) | Inf time (fps) | AP@0.25 | AP@0.5 | Download |
|---|---|---|---|---|---|
| MinkResNet34 | 10.5 | 15.7 | 69.7(70.7*) | 55.2(56.0*) | model | log |
SUN RGB-D
| Backbone | Mem (GB) | Inf time (fps) | AP@0.25 | AP@0.5 | Download |
|---|---|---|---|---|---|
| MinkResNet34 | 6.3 | 17.9 | 63.8(63.8*) | 47.3(48.2*) | model | log |
S3DIS
| Backbone | Mem (GB) | Inf time (fps) | AP@0.25 | AP@0.5 | Download |
|---|---|---|---|---|---|
| MinkResNet34 | 23.5 | 10.9 | 67.4(64.9*) | 45.7(43.8*) | model | log |
Note
- We report the results across 5 train runs followed by 5 test runs. * means the results reported in the paper.
- Inference time is given for a single NVidia RTX 4090 GPU. All models are trained on 2 GPUs.
Citation
@inproceedings{rukhovich2022fcaf3d,
title={FCAF3D: Fully Convolutional Anchor-Free 3D Object Detection},
author={Danila Rukhovich, Anna Vorontsova, Anton Konushin},
booktitle={European conference on computer vision},
year={2022}
}