DSNet：A Novel Way to Use Atrous Convolutions in Semantic Segmentation

a novel real-time model in semantic segmentation. Paper: ([PDF]https://arxiv.org/abs/2406.03702))

This is the implementation for DSNet. This paper was accepted by CVPR 2024 Workshop on Computer Vision in the Wild. DSNetV2 is currently under development... DSNetV2 will offer a better balance between speed and accuracy, as well as a more comprehensive DSNet family ranging from small to large models.

Environment:

PyTorch 1.10

python 3.8

4*RTX4090 or 8*RTX4090

  pip install -r requirements.txt

Highlight

• We revisited the design of atrous convolutions in CNNs,and explored three empirical guidelines for atrous convolution. Based on the above guidelines, we proposed a novel Dual-branch network.

• DSNet achieves a new state-of-the-art trade-off between accuracy and speed on ADE20K, Cityscapes,and BDD10K.

Overview:

overview-of-our-method
An overview of the basic architecture of our proposed DSNet.

Diagram of Multi-Scale Fusion Atrous Convolutional Block (MFACB).

Train and Inference speed:

This implementation is based on HRNet-Semantic-Segmentation and PIDNet. Please refer to their repository for installation and dataset preparation. The inference speed is tested on single RTX 3090 or RTX4090. BDD10K has not been implemented in the above link. The dataset storage format is as follows. Download link: web page

bdd
- seg
- color_labels
  - train
  - val
- images
  - train
  - val
  - test
- labels
  - train
  - val

Train

python -m torch.distributed.launch --nproc_per_node=4 DSNet/tools/train.py

Inference speed

python DSNet/models/speed/dsnet_speed.py

Citation

If you think this work is useful, please cite our paper:

@misc{guo2024dsnet,
      title={DSNet: A Novel Way to Use Atrous Convolutions in Semantic Segmentation}, 
      author={Zilu Guo and Liuyang Bian and Xuan Huang and Hu Wei and Jingyu Li and Huasheng Ni},
      year={2024},
      eprint={2406.03702},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}