Lightweight Context-Aware Network Using Partial-Channel Transformation for Real-Time Semantic Segmentation

Lightweight Context-Aware Network Using Partial-Channel Transformation for Real-Time Semantic Segmentation

Segmentation performance of LCNet

S1	S2	Crop Size*	Dataset	Pretrained	Train type	mIoU	Params	Speed	Location
3	7	512,1024	Cityscapes	No	trainval	73.3	0.51	185	-
3	7	1024,1024	Cityscapes	No	trainval	73.8	0.51	142	-
3	11	512,1024	Cityscapes	No	trainval	74.3	0.74	136	-
3	11	1024,1024	Cityscapes	No	train	75.6	0.74	117	-
3	11	1024,1024	Cityscapes	No	trainval	75.8	0.74	117	-

* Represents the resolution of the input image cropping in the training phase.

Preparation

You need to download the Cityscapes and CamVid datasets and place the symbolic links or datasets of the Cityscapes and CamVid datasets in the dataset directory. Our file directory is consistent with DABNet (https://github.com/Reagan1311/DABNet).

dataset
  ├── camvid
  |    ├── train
  |    ├── test
  |    ├── val 
  |    ├── trainannot
  |    ├── testannot
  |    ├── valannot
  |    ├── camvid_trainval_list.txt
  |    ├── camvid_train_list.txt
  |    ├── camvid_test_list.txt
  |    └── camvid_val_list.txt
  ├── cityscapes
  |    ├── gtCoarse
  |    ├── gtFine
  |    ├── leftImg8bit
  |    ├── cityscapes_trainval_list.txt
  |    ├── cityscapes_train_list.txt
  |    ├── cityscapes_test_list.txt
  |    └── cityscapes_val_list.txt           

How to run

1 Training

1.1 Cityscapes

python train.py

1.2 CamVid

python train.py --dataset camvid --train_type trainval --max_epochs 1000 --lr 1e-3 --input_size 360,480

2 Testing

python test.py --dataset ${camvid, cityscapes} --checkpoint ${CHECKPOINT_FILE}

2.1 Cityscapes

python test.py --dataset cityscapes --checkpoint "./checkpoints/LCNet_3_11_1024_train.pth"

To convert the training lables to class lables.

python trainID2labelID.py Package the file into xxx.zip Submit the zip file to https://www.cityscapes-dataset.com/submit/. You can get the results from the https://www.cityscapes-dataset.com/submit/.

2.2 CamVid

python test.py --dataset camvid --checkpoint ${CHECKPOINT_FILE}

3. fps

python eval_forward_time.py --size 512,1024

Citation

@ARTICLE{
  10411824,
  author={Shi, Min and Lin, Shaowen and Yi, Qingming and Weng, Jian and Luo, Aiwen and Zhou, Yicong},
  journal={IEEE Transactions on Intelligent Transportation Systems},
  title={Lightweight Context-Aware Network Using Partial-Channel Transformation for Real-Time Semantic Segmentation},
  year={2024},
  volume={},
  number={},
  pages={1-16}
}

Reference

https://github.com/xiaoyufenfei/Efficient-Segmentation-Networks

https://github.com/Reagan1311/DABNet