islamamirul / rsdnet

Saliency Ranking

Revisiting Salient Object Detection: Simultaneous Detection, Ranking, and Subitizing of Multiple Salient Objects

This repository contains code for the paper

Revisiting Salient Object Detection: Simultaneous Detection, Ranking, and Subitizing of Multiple Salient Objects,
Presented at CVPR 2018

If you find the code useful for your research, please consider citing our work:

@InProceedings{Islam_2018_CVPR,
  author = {Amirul Islam, Md and Kalash, Mahmoud and Bruce, Neil D. B.},
  title = {Revisiting Salient Object Detection: Simultaneous Detection, Ranking, and Subitizing of Multiple Salient Objects},
  booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  month = {June},
  year = {2018}
}

How to use these codes?

Anyone can freely use our codes for what-so-ever purpose they want to use. Here we give a detailed instruction to set them up and use for different applications.

The codes can be downloaded using the following command:

    git clone --recursive https://github.com/islamamirul/rsdnet.git
    cd rsdnet

Setup:

1. Download and compile caffe-rsdnet which is a modified version of deeplab-public-ver2

2. Download the PASCAL-S dataset from here and put them under ./data/

3. Run the following script to generate stack of saliency masks

    ./scripts/stack_generation/generate_saliency_mask_stack.m
   

4. Download the pretrained-weights (init.caffemodel) of DeepLabv2 from here and put it under ./models/trained_weights/

Trained Model

You can download the trained model which is reported in our paper at Dropbox and put them under ./models/trained_weights/

Training RSDNet

Modify the caffe root directory and run the following command to start training:

sh train_rsdnet.sh

Testing RSDNet

Modify the caffe root directory in ./scipts/inference/test_rsdnet.py and run the following command:

sh test_rsdnet.sh

Results

The results of multiple salienct object detection (extended to salient object ranking) on PASCAL-S dataset can be found at Dropbox

Salient Object Ranking (SOR): Please run the following script to generate overall SOR score for RSDNet

       ./scripts/eval/SOR/SOR.m

To generate the detection scores (F-measure, AUC, and MAE) please run the corresponding scripts under .scripts/eval/..