DSLRDNet: Addressing Multiple Salient Object Detection via Dual-Space Long-Range Dependencies
Prerequisites:
- Pytorch 1.2.0
- Opencv 2.4.5
- TensorboardX
For training:
- Download the DUTS-TR (Google Drive) training dataset.
- Download the initial pratrained VGG/ResNet (Google Drive) model.
- Change the training data path in dataset.py.
- Change the training settings in solver.py and run.py
- Start to train with
python3 run.py --mode train
For testing:
- Download the pretrained models (UoN server).
- Change the data path in dataset.py
- Change the test settings in run.py.
- Generate saliency maps with
python3 run.py --mode test --sal_mode m
, where 'm' demonstrates the MSOD dataset. - We use the public open source evaluation code. (https://github.com/weijun88/F3Net)
Datasets and results: