LabelDistill: Label-guided Cross-modal Knowledge Distillation for Camera-based 3D Object Detection
Sanmin Kim, Youngseok Kim, Sihwan Hwang, Hyeonjun Jeong, and Dongsuk Kum, ECCV 2024

Recent advancements in camera-based 3D object detection have introduced cross-modal knowledge distillation to bridge the performance gap with LiDAR 3D detectors, leveraging the precise geometric information in LiDAR point clouds. However, existing cross-modal knowledge distillation methods tend to overlook the inherent imperfections of LiDAR, such as the ambiguity of measurements on distant or occluded objects, which should not be transferred to the image detector. To mitigate these imperfections in LiDAR teacher, we propose a novel method that leverages aleatoric uncertainty-free features from ground truth labels. In contrast to conventional label guidance approaches, we approximate the inverse function of the teacher's head to effectively embed label inputs into feature space. This approach provides additional accurate guidance alongside LiDAR teacher, thereby boosting the performance of the image detector. Additionally, we introduce feature partitioning, which effectively transfers knowledge from the teacher modality while preserving the distinctive features of the student, thereby maximizing the potential of both modalities. Experimental results demonstrate that our approach improves mAP and NDS by 5.1 points and 4.9 points compared to the baseline model, proving the effectiveness of our approach.

1. Create a conda virtual environment

conda create --name labeldistill python=3.8 -y
conda activate labeldistill

2. Install PyTorch (v1.9.0)

pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 torchaudio==0.9.0 -f https://download.pytorch.org/whl/torch_stable.html

3. Clone repo

git clone https://github.com/sanmin0312/LabelDistill.git

4. Install mmcv, mmdet and mmseg

pip install openmim
mim install mmcv-full==1.6.0
mim install mmdet==2.26.0
mim install mmsegmentation==0.29.1

5. Install mmdet3d

cd LabelDistill
git clone -b v1.0.0rc4 https://github.com/open-mmlab/mmdetection3d.git
cd mmdetection3d
pip install -e .

6. Install requirements

cd ..
pip install -r requirements.txt
python setup.py develop

1. Downlaod nuScenes official dataset & make symlink

ln -s [nuscenes root] ./data/

2. Prepare infos

python scripts/gen_info.py

3. Generate lidar depth

python scripts/gen_depth_gt.py

The directory should be as follows.

LabelDistill
├── data
│   ├── nuScenes
│   │   ├── nuscenes_infos_train.pkl
│   │   ├── nuscenes_infos_val.pkl
│   │   ├── maps
│   │   ├── samples
│   │   ├── depth_gt
│   │   ├── sweeps
│   │   ├── v1.0-trainval

Training

python [EXP_PATH] --amp_backend native -b 4 --gpus 4

Evaluation

python [EXP_PATH] --ckpt_path [CKPT_PATH] -e -b 4 --gpus 4

@inproceedings{kim2024labeldistill,
  title={LabelDistill: Label-guided Cross-modal Knowledge Distillation for Camera-based 3D Object Detection},
  author={Kim, Sanmin and Kim, Youngseok and Hwang, Sihwan and Jeong, Hyeonjun and Kum, Dongsuk},
  booktitle={Proceedings of the European Conference on Computer Vision},
  year={2024}
}

• BEVDepth
• mmdet3d
• CenterPoint