Skip to content

Latest commit

 

History

History
171 lines (134 loc) · 4.96 KB

README.md

File metadata and controls

171 lines (134 loc) · 4.96 KB
Raw

T-RODNet: Transformer for Vehicular Millimeter-Wave Radar Object Detection

T-RODNet has been accepted by IEEE Transactions on Instrumentation and Measurement (TIM).

T-RODNet Overview

Please cite our paper if this repository is helpful for your research:

Statement: This code is based on RODNet.[Arxiv]

Datasets

CRUW

CARRADA

Results

On the CRUW dataset

Models AP AR
RODNet-CDC 76.33 79.28
RODNet-HG 79.43 83.59
RODNet-HWGI 78.06 81.07
DCSN 75.30 79.92
T-RODNet 83.27 86.98

On the CARRADA dataset

Models mIoU mDice
FCN-8s 34.5 40.9
U-Net 32.8 38.2
DeepLabv3+ 32.7 38.3
RSS-Net 32.1 37.8
RAMP-CNN 27.9 30.5
MV-Net 26.8 28.5
TMVA-Net 41.3 51.0
T-RODNet 43.5 53.6

Installation

Create a conda environment for T-RODNet. Tested under Python 3.6, 3.7, 3.8.

conda create -n trodnet python=3.* -y
conda activate trodnet

Install pytorch.
Note: If you are using Temporal Deformable Convolution (TDC), we only tested under pytorch<=1.4 and CUDA=10.1. Without TDC, you should be able to choose the latest versions. If you met some issues with environment, feel free to raise an issue.

conda install pytorch=1.4 torchvision cudatoolkit=10.1 -c pytorch  # if using TDC
# OR
conda install pytorch torchvision cudatoolkit=10.1 -c pytorch  # if not using TDC

Setup RODNet package.

pip install -e .

Note: If you are not using TDC, you can rename script setup_wo_tdc.py as setup.py, and run the above command. This should allow you to use the latest cuda and pytorch version.

Evaluation

If you want to installcruw-devkit package, please refer to cruw-devit repository for detailed instructions.

Note: Our code have revised the 'cruw' evaluation package. You can use it in (evaluate/ex_evaluate_rod2021.py).

Prepare data for RODNet

Download ROD2021 dataset. Follow this script to reorganize files as below.

Note: To facilitate testing and evaluation of network performance. We strongly recommend testing as follows:

The test set uses the following 4 sequences from the training set:
[2019_04_09_BMS1001, 2019_04_30_MLMS001, 2019_05_23_PM1S013, 2019_09_29_ONRD005].

The train set uses the remaining 36 sequences.

data_root
  - sequences
  | - train---------------------------------------> 36 train set 
  | | - <SEQ_NAME>
  | | | - IMAGES_0
  | | | | - <FRAME_ID>.jpg
  | | | | - ***.jpg
  | | | - RADAR_RA_H
  | | |   - <FRAME_ID>_<CHIRP_ID>.npy
  | | |   - ***.npy
  | | - ***
  | | 
  | - test----------------------------------------> 4 test set 
  |   - <SEQ_NAME>
  |   | - RADAR_RA_H
  |   |   - <FRAME_ID>_<CHIRP_ID>.npy
  |   |   - ***.npy
  |   - ***
  | 
  - annotations
  | - train---------------------------------------> 36 train set
  | | - <SEQ_NAME>.txt
  | | - ***.txt
  | - test----------------------------------------> 4 test set 
  |   - <SEQ_NAME>.txt
  |   - ***.txt
  - calib

Convert data and annotations to .pkl files.

python tools/prepare_dataset/prepare_data.py \
        --config configs/<CONFIG_FILE> \
        --data_root <DATASET_ROOT> \
        --split train,test \
        --out_data_dir data/<DATA_FOLDER_NAME>

Train models

python tools/train.py --config configs/<CONFIG_FILE> \
        --data_dir data/<DATA_FOLDER_NAME> \
        --log_dir checkpoints/

Inference

Note: Please change the 'out_path' in (test.py).

python tools/test.py --config configs/<CONFIG_FILE> \
        --data_dir data/<DATA_FOLDER_NAME> \
        --checkpoint <CHECKPOINT_PATH> \
        --res_dir results/

Evaluate models

In this section, we have created new evaluation tools to facilitate testing. Therefore there is no need to test by uploading the website.

python evaluate/ex_evaluate_rod2021.py

Citation

If you find this article very helpful in your research, or if you wish to have a reference when using our results, please cite the following papers:

@ARTICLE{9989400,
  author={Jiang, Tiezhen and Zhuang, Long and An, Qi and Wang, Jianhua and Xiao, Kai and Wang, Anqi},
  journal={IEEE Transactions on Instrumentation and Measurement}, 
  title={T-RODNet: Transformer for Vehicular Millimeter-Wave Radar Object Detection}, 
  year={2023},
  volume={72},
  number={},
  pages={1-12},
  doi={10.1109/TIM.2022.3229703}}