yowenyo / fedbevt

FL Transformer combine.

FedBEVT: Federated Learning Bird’s Eye View Perception Transformer in Road Traffic Systems

FedBEVT is a federated learning framework to train a transformer-based model for road environmental BEV perception.

Installation

List of dependencies

• matplotlib
• PyTorch (1.8.1)
• numpy
• open3d
• opencv-python
• tensorboardX
• shapely
• einops
• cumm
• Cython
• timm

Install requirements

cd fedbevt

# Initialize conda env
conda create -y --name fedbevt python=3.8
conda activate fedbevt
conda install pytorch==1.11.0 torchvision==0.12.0 cudatoolkit=11.3 -c pytorch

# Install dependencies
python fedbevt/utils/setup.py build_ext --inplace
python setup.py develop

Federated Data Preparation

Download

We follow the OPV2V data format, extending orignal OPV2V data (Car) with images and BEV lables for Trucks, Bus and Infrastructure: DOWNLOAD.

More details on OPV2V data format can be found HERE.

Preprocess

The data should be distributed to different clients according use cases. We provide the script and corresponding YAML-files for each use case.

Example for use case 1: python3 fedbevt/tools/fed_data_utils.py --config uc1.yaml --source /data/path/to/data_resource --target /data/fedbevt/uc1

Explanations of arguments:

• --config: The configuration file in YAML.
• --source: Source directory to original dataset.
• --target: Target directory to federated dataset.

Execution FedBEVT

Example: python fedbevt/fed/fedbevt_main.py --config fed_fax_flserver_uc1.yaml --gpu 0 --per avg

Explanations of arguments:

• --config: The configuration file in YAML
• --gpu: The id of GPU
• --per: The approach used for personalization: 'avg' and 'cap' are supported

Evaluation Procedures

Run python fedbevt/visualization/eval_main.py -rr 'results/json_logs' to plot the testing accuracy and training loss by the increasing number of epochs or communication rounds. Note that the labels in the figure is the name of result files

Hyperparameters and Commands

We provide tables of our recommended hyperparameters and commands

Data preparation for UC1-4

UC ID	Command
#1	python fedbevt/tools/fed_data_utils.py --config uc1.yaml --source /data/path/to/data_resource --target /data/fedbevt/uc1
#2	python fedbevt/tools/fed_data_utils.py --config uc2.yaml --source /data/path/to/data_resource --target /data/fedbevt/uc2
#3	python fedbevt/tools/fed_data_utils.py --config uc3.yaml --source /data/path/to/data_resource --target /data/fedbevt/uc3
#4	python fedbevt/tools/fed_data_utils.py --config uc4.yaml --source /data/path/to/data_resource --target /data/fedbevt/uc4

Recommended hyperparameters

Training parameters:

Parameter	batch size	local epoches	num of rounds	eval freq	save freq	max cav num
Value	16	1	500	5	5	1

Optimizer parameters:

Parameter	core method	learning rate	eps	weight decay
Value	AdamW	2e-4	1e-10	1e-2

Learning rate scheduler parameters:

Parameter	core method	warmup learning rate	warmup round	min. learning rate
Value	cosineannealwarm	2e-5	10	5e-6

Experiment commands

UC1:

Exp ID	--config	--per
#1	fed_fax_flserver_uc1.yaml	avg
#2	fed_fax_flserver_uc1.yaml	cap

UC2:

Exp ID	--config	--per
#3	fed_fax_flserver_uc2.yaml	avg
#4	fed_fax_flserver_uc2.yaml	cap

UC3:

Exp ID	--config	--per
#5	fed_fax_flserver_uc3.yaml	avg
#6	fed_fax_flserver_uc3.yaml	cap

UC4:

Exp ID	--config	--per
#7	fed_fax_flserver_uc4.yaml	avg
#8	fed_fax_flserver_uc4.yaml	cap

Citation

If you find FedBEVT is useful, please cite our work:

@article{song2023fedbevt,
  title={FedBEVT: Federated Learning Bird's Eye View Perception Transformer in Road Traffic Systems},
  author={Song, Rui and Xu, Runsheng and Festag, Andreas and Ma, Jiaqi and Knoll, Alois},
  journal={arXiv preprint arXiv:2304.01534},
  year={2023}
}

Acknowledgement

The project extensively utilizes the toolchains in the OpenCDA ecosystem, including the OpenCOOD and the OpenCDA simulation tools.