In this project, we build a simple federated deep learning model (i.e., LeNet model) that classifies 62 differents Belgian Traffic Signs.
- You can download dataset from here.
- You also can download dataset from Google Drive.
- Baidu Yun, pwd: wqtv.
In this project, we use LeNet model to classify 62 differents Belgian Traffic Signs. For more detials can be seen in [1].
[1]: In Proceedings of the IEEE [bibtex]
@ARTICLE{726791,
author={Y. {Lecun} and L. {Bottou} and Y. {Bengio} and P. {Haffner}},
journal={Proceedings of the IEEE},
title={Gradient-based learning applied to document recognition},
year={1998},
volume={86},
number={11},
pages={2278-2324},
doi={10.1109/5.726791}}
Code: PyTorch implementation of LeNet-5 with live visualization.
Please note that you can replace other convolutional neural networks such as ResNet and AlexNet model. You just need to pay attention to the input size of different convolutional neural networks.
- Setp 1: Download the dataset via the above links.
- Setp 2: If you want to run the centralized LeNet model, please via the following command line:
python main_nn.py --dataset traffic --iid --num_channels 3 --model LeNet --epochs 500 --gpu -1
- If you want to run the federated LeNet model, please via the following command line:
python main_fed.py --dataset traffic --iid --num_channels 3 --model LeNet --epochs 1000 --gpu -1
For the following command:
dataset_train, dataset_test = get_train_valid_loader('/home/liuyi/Documents/federated-learning-master/federated-learning-master/data', batch_size=32, num_workers=0)
You need to use your data path.
- Setp 3: For centralized learning, you can access the folder log/.... For federated learning, you can access the folder save/....
