ExtractingBuildingByMMSeg
This repository is for Competition for ML_data class. Based on mmsegmentatoin,mainly using swin transformer to completed the competition.
Two ways of training are provided here, non-distributed and distributed training. If you want to use non-distributed training, you can directly check the code in test.py to train and predict.
If you want to use distributed training on a linux server, you can do the following： First make dist_train.sh available under linux
chmod 777 ./mmsegmentation/tools/dist_train.sh
vi ./mmsegmentation/tools/dist_train.sh
set ff=unix
Next, you can use the following commands for distributed training. We provide two kinds of configuration files, biet and swin, so you can configure the configuration files yourself as needed, and to support both BEiT and ConvNeXt networks, we modify and add to the mmsegmentation source code to make sure it can be used directly.
nohup ./mmsegmentation/tools/dist_train.sh ./mine/myconfig_swin.py 4 > hehe.log 2>&1 &
You can also manually select the desired GPU
CUDA_VISIBLE_DEVICES=2,3 ./mmsegmentation/tools/dist_train.sh ./mine/myconfig_biet.py 2
For files, you can define your own file types and add them to /mmsegmentation/mmseg/datasets/ in the following way.
import os.path as osp
from .builder import DATASETS
from .custom import CustomDataset
classes = ('buildinding', 'background')
palette = [[0, 0, 0], [255, 255, 255]]
@DATASETS.register_module()
class myDataset(CustomDataset):
CLASSES = classes
PALETTE = palette
def __init__(self, split, **kwargs):
super().__init__(img_suffix='.tif', seg_map_suffix='.tif',
split=split, **kwargs)
assert osp.exists(self.img_dir) and self.split is not None
After getting the pth prediction, you can use the predict.py to code the prediction and output the result.
29 Aug 22, 2022
70 Oct 29, 2022
9 Dec 1, 2022
249 Jan 3, 2023
266 Nov 24, 2022
37 Oct 30, 2022
24 Sep 8, 2022
68 Dec 9, 2022
40 Dec 22, 2022
258 Dec 31, 2022
4 Apr 15, 2022
104 Nov 22, 2022
62 Dec 28, 2022
69 Sep 29, 2022
298 Dec 21, 2022
94 Dec 25, 2022
45 Aug 31, 2022
10 Jul 14, 2022
20 Dec 9, 2021