Anchor-x/fed-yolo
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

更新配置文件，移除 YOLO 版本实现计划，调整 UAV 和 COCO 配置中的数据量和批次大小

Browse Source
This commit is contained in:
Yunhao Meng
2025-10-31 13:14:47 +08:00
parent 291b82bec3
commit fdb70869f9
3 changed files with 10 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
README.md
View File

@@ -19,9 +19,6 @@ nohup bash fed_run.sh 1 > train.log 2>&1 &
- Implement FedProx - Implement FedProx
- Implement SCAFFOLD - Implement SCAFFOLD
- Implement FedNova - Implement FedNova
- Add more YOLO versions (e.g., YOLOv8, YOLOv5, etc.)
- Implement YOLOv8
- Implement YOLOv5
# references # references
[PyTorch Federated Learning](https://github.com/rruisong/pytorch_federated_learning) [PyTorch Federated Learning](https://github.com/rruisong/pytorch_federated_learning)

4
config/coco_cfg.yaml
View File

@@ -17,8 +17,8 @@ local_batch_size: 32 # local training batch size
val_batch_size: 128 # validation batch size val_batch_size: 128 # validation batch size
num_workers: 8 # number of data loader workers num_workers: 8 # number of data loader workers
min_data: 1700 # minimum number of images per client min_data: 1800 # minimum number of images per client
max_data: 1800 # maximum number of images per client max_data: 1900 # maximum number of images per client
partition_mode: "overlap" # "overlap" or "disjoint" partition_mode: "overlap" # "overlap" or "disjoint"
connection_ratio: 1 # connection ratio, e.g., 1.0 means all clients connection_ratio: 1 # connection ratio, e.g., 1.0 means all clients

16
config/uav_cfg.yaml
View File

@@ -3,22 +3,22 @@ fed_algo: "FedAvg" # federated learning algorithm
model_name: "yolo_v11_n" # yolo_v11_n, yolo_v11_t, yolo_v11_s, yolo_v11_m, yolo_v11_l, yolo_v11_x model_name: "yolo_v11_n" # yolo_v11_n, yolo_v11_t, yolo_v11_s, yolo_v11_m, yolo_v11_l, yolo_v11_x
i_seed: 202509 # initial random seed i_seed: 202509 # initial random seed
num_client: 100 # total number of clients num_client: 36 # total number of clients
num_round: 500 # total number of communication rounds num_round: 50 # total number of communication rounds
num_local_class: 1 # number of classes per client num_local_class: 1 # number of classes per client
res_root: "results" # root directory for results res_root: "results" # root directory for results
dataset_path: "/home/image1325/ssd1/dataset/uav/" dataset_path: "/mnt/DATA/uav/"
# train_txt: "train.txt" # path to training set txt file # train_txt: "train.txt" # path to training set txt file
# val_txt: "val.txt" # path to validation set txt file # val_txt: "val.txt" # path to validation set txt file
# test_txt: "test.txt" # path to test set txt file # test_txt: "test.txt" # path to test set txt file
local_batch_size: 32 # local training batch size local_batch_size: 36 # local training batch size
val_batch_size: 16 # validation batch size val_batch_size: 128 # validation batch size
num_workers: 4 # number of data loader workers num_workers: 8 # number of data loader workers
min_data: 640 # minimum number of images per client min_data: 385 # minimum number of images per client
max_data: 720 # maximum number of images per client max_data: 400 # maximum number of images per client
partition_mode: "overlap" # "overlap" or "disjoint" partition_mode: "overlap" # "overlap" or "disjoint"
connection_ratio: 1 # connection ratio, e.g., 1.0 means all clients connection_ratio: 1 # connection ratio, e.g., 1.0 means all clients