Code-A Novel Attention-Based Network for Fast Salient Object Detection

Our Improved EGNet

For training:

1. Clone this code by git clone https://gitee.com/binzhangbinzhangbin/code-a-novel-attention-based-network-for-fast-salient-object-detection.gits, assume your source code directory is$EGNet;

2. Download training data (fsex) (google drive);

3. Download initial model (8ir7) (google_drive);

4. Change the image path and intial model path in run.py and dataset.py;

5. Start to train with python3 run.py --mode train.

For testing:

1. Download pretrained model (2cf5) (google drive);

2. Change the test image path in dataset.py

3. Generate saliency maps for SOD dataset by python3 run.py --mode test --sal_mode s, PASCALS by python3 run.py --mode test --sal_mode p and so on;

4. Testing code we use is the public open source code. (https://github.com/Andrew-Qibin/SalMetric)

Pretrained models, datasets and results:

| Page |