CTDNet

The PyTorch code for ACM MM2021 paper "Complementary Trilateral Decoder for Fast and Accurate Salient Object Detection"

Requirements

• Python 3.6
• Pytorch 1.0+
• OpenCV 4.0
• Numpy
• TensorboardX
• Apex

Dataset

Download the SOD datasets and unzip them into data folder.

Model

• If you want to test our method, please download the model into out folder.
• If you want to train our method, please download the pretrained model into res folder.

Train

python train.py

• We implement our method by PyTorch and conduct experiments on a NVIDIA 1080Ti GPU.
• We adopt ResNet-18 and ResNet-50 pre-trained on ImageNet as backbone networks, respectively.
• We train our method on DUTS-TR and test our method on other datasets.
• After training, the result models will be saved in out folder.

Test

python test.py

• After testing, saliency maps will be saved in eval folder.

Results

Evaluation

Reference

This project is based on the implementation of F3Net.