SpiderMesh: Spatial-aware Demand-guided Recursive Meshing for RGB-T Semantic Segmentation

For technical details, please refer to:

SpiderMesh: Spatial-aware Demand-guided Recursive Meshing for RGB-T Semantic Segmentation

(0) Abstract

For semantic segmentation in urban scene understanding, RGB cameras alone often fail to capture a clear holistic topology, especially in challenging lighting conditions. Thermal signal is an informative additional channel that can bring to light the contour and fine-grained texture of blurred regions in low-quality RGB image. Aiming at RGB-T (thermal) segmentation, existing methods either use simple passive channel/spatial-wise fusion for cross-modal interaction, or rely on heavy labeling of ambiguous boundaries for fine-grained supervision. We propose a Spatial-aware Demand-guided Recursive Meshing (SpiderMesh) framework that: 1) proactively compensates inadequate contextual semantics in optically-impaired regions via a demand-guided target masking algorithm; 2) refines multimodal semantic features with recursive meshing to improve pixel-level semantic analysis performance. We further introduce an asymmetric data augmentation technique M-CutOut, and enable semi-supervised learning to fully utilize RGB-T labels only sparsely available in practical use. Extensive experiments on MFNet and PST900 datasets demonstrate that SpiderMesh achieves new state-of-the-art performance on standard RGB-T segmentation benchmarks.

(1) Setup

This code has been tested with Python 3.8.10, Pytorch 1.11.0, CUDA 11.7.

• Setup environment

  conda create -n SpiderMesh python=3.8.10
  source activate SpiderMesh

• Clone the respository

• Install the requirements

  pip install -r requirements.txt
  pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu113
  

• Download dataset

  Dataset	Note	Link
  MFNet	preprocessed by RTFNet*	Download:5ah9
  PST900	original dataset	Download

  The original link of the preprocessed MFNet (by RTFNet) seems not availuable now, we have upload it to our BaiduCloud.

• Download pretrained models

  Dataset	Description	Link
  MFNet	SpiderMesh-50	Download:btkl
  	SpiderMesh-101	Download:5t2p
  	SpiderMesh-152	Download:i3me
  	Semi-supervised SpiderMesh-152	Download:8jif
  PST900	SpiderMesh-152	Download:5qzy

(2) Training

• Supervised learning

  python train.py # for MFNet dataset
  python train_pst.py # for PST900 dataset

  

• Semi-supervised learning

  python semi-train.py

  

(3) Evaluation

For detailed analysis, three modes are supported: test(default), test_day, test_night.

python test.py # for MFNet dataset
python test_pst.py # for PST900 dataset

• Complexity

Citation

If you find our work useful in your research, please consider citing:

@article{spidermesh,
  title={{SpiderMesh}: Spatial-aware Demand-guided Recursive Meshing for RGB-T Semantic Segmentation},
  author={S. Fan and Z. Wang and Y. Wang and J. Liu},
  journal={arXiv:2303.08692},
  year={2023}}