RGB-T-segmentation

Datasets for RGB-T SOD

• We follow the work 'LSNet: Lightweight Spatial Boosting Network for Detecting Salient Objects in RGB-Thermal Images', IEEE TIP, 2023.(https://github.com/zyrant/LSNet)

• Training and Testing Dataset: https://drive.google.com/file/d/1vjdD13DTh9mM69mRRRdFBbpWbmj6MSKj/view

Requirement

• Python 3.7
• PyTorch 1.13.1
• torchvision
• numpy
• Pillow
• Cython
• mmcv

Training Model on VT5000

Fully supervised Training

• 1.Run train_sod.py

Testing on VT5000,VT1000,VT821

• 1.Set the path of testing sets in test_sod.py
• 2.Run test_sod.py (can generate the segmentation maps) : python test_sod.py --exp_name='(checkpoint folder)'
• 3.Run test_score.py (change root, method_name), record MAE, E, F ,S.