TiGDistill-BEV: Multi-view BEV 3D Object Detection via Target Inner-Geometry Learning Distillation

Paper

Official implementation of "TiGDistill-BEV: Multi-view BEV 3D Object Detection via Target Inner-Geometry Learning Distillation".

Introduction

We introduce the TiGDistill-BEV, a novel approach that bridges this gap by distilling the Target Inner-Geometry learning scheme to enhance camera-based BEV detectors from both depth and BEV feature by leveraging the diverse modalities. We propose two key modules: an inner-depth supervision module to learn the low-level relative depth relations within objects which equips detectors with a deeper understanding of object-level spatial structures, and an inner-feature BEV distillation module to transfer high-level semantics of different keypoints within foreground targets. To further alleviate the domain gap, we incorporate both inter-channel and inter-keypoint distillation to model feature similarity. Extensive experiments on the nuScenes benchmark demonstrate that TiGDistillBEV significantly boosts camera-based detectors achieving a state-of-the-art with 62.8% NDS.

Main Results

Method	mAP	NDS
TiGDistill-BEV-R50	33.8	37.5
TiGDistill-BEV4D-R50	36.6	46.1
TiGDistill-PillarNext-R50-CBGS	38.7	50.4
TiGDistill-BEVFusion-R50-CBGS	39.6	51.1

We provide the model and log of TiG-BEV4D-R101-CBGS.

Method	mAP	NDS	Model	Log
TiGDistill-BEV4D-R101-CBGS	44.0	54.4	Google	Google

Quick Start

Installation

Please see getting_started.md in BEVDet.

Data Preparation

nuScenes

Please follow the instructions from here to download and preprocess the nuScenes dataset. Please remember to download both detection dataset and the map extension (for BEV map segmentation). After data preparation, you will be able to see the following directory structure (as is indicated in mmdetection3d):

TiGDistill-BEV
├── mmdet3d
├── tools
├── configs
├── data
│   ├── nuscenes
│   │   ├── maps
│   │   ├── samples
│   │   ├── sweeps
│   │   ├── v1.0-test
|   |   ├── v1.0-trainval
│   │   ├── nuscenes_database
│   │   ├── nuscenes_infos_train.pkl
│   │   ├── nuscenes_infos_val.pkl
│   │   ├── nuscenes_infos_test.pkl
│   │   ├── nuscenes_dbinfos_train.pkl

Train & Evaluate in Command Line

Now we only support training and evaluation with gpu. Cpu only mode is not supported.

Use the following command to start a distributed training using 4 GPUs. The models and logs will be saved to work_dirs/CONFIG_NAME

python -m torch.distributed.launch --nproc_per_node=4 ./tools/train.py CONFIG_PATH

For distributed testing with 4 gpus,

bash dist_train.sh CONFIG_PATH 4 
or 
python -m torch.distributed.launch --nproc_per_node=4 ./tools/dist_test.py CONFIG_PATH --work_dir work_dirs/CONFIG_NAME --checkpoint work_dirs/CONFIG_NAME/latest.pth 

For testing with one gpu and see the inference time,

bash dist_train.sh CONFIG_PATH  work_dirs/CONFIG_NAME/latest.pth 4 
or 
python ./tools/dist_test.py CONFIG_PATH work_dirs/CONFIG_NAME/latest.pth --work_dir work_dirs/CONFIG_NAME 

Acknowledgement

We sincerely thank these great open-sourced work below:

• open-mmlab
• CenterPoint
• Lift-Splat-Shoot
• BEVDet
• BEVDepth
• BEVFusion