ITSD-pytorch

Code for CVPR 2020 paper "Interactive Two-Stream Decoder for Accurate and Fast Saliency Detection"

Saliency maps can be download at: VGG (Baidu Yun [gf1i]), Resnet (Baidu Yun [sanf])

This code is somehow outdated, please use the implementation in our SOD benchmark.

We release our new works on Unsupervised Salient Object Detection (USOD) at A2S-USOD and A2S-v2.

Prerequisites

• Pytorch 1.0.0
• torchvision 0.2.1
• Thop
• Progress

Usage：

Official imagenet-pretrained weights can be download at Resnet50 and VGG16.

Our models: Google drive (ResNet50 and VGG16) or Baidu Disk (Resnet50 [y55w] and VGG16 [kehh])

Please refer to this repo for results evaluation: SalMetric.

Training：

python3 train.py --sub=[job_name] --ids=[gpus] --model=[vgg/resnet]

Testing:

mv path_to_model ./save/[vgg/resnet]/[job_name]/final.pkl  # if testing the provided models
python3 test.py --sub=[job_name] --ids=[gpus] --model=[vgg/resnet]

Contact

If you have any question, feel free to contact me via: mootheszhou@gmail.com.

Bibtex

@InProceedings{Zhou_2020_CVPR,
author = {Zhou, Huajun and Xie, Xiaohua and Lai, Jian-Huang and Chen, Zixuan and Yang, Lingxiao},
title = {Interactive Two-Stream Decoder for Accurate and Fast Saliency Detection},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2020}
}