Generate Saliency maps for Pyramid Feature Attention Network for Saliency Detection, CVPR 2019

Install Dependencies

The code is written in Python 3.6 using the following libraries:

numpy
tqdm
opencv-python
torch==1.1.0
torchvision==0.3.0

Install the libraries using requirements.txt as:

pip install -r requirements.txt

Data

We use ImageNet dataset.

Pre-trained Model

Download the pre-trained model from Google Drive.

Usage

Use the command below to generate and save saliency maps with different thresholds:

python inference.py --save_img --sal_threshold 0.1

Reference

Keras implementation : LINK