MMFT

Joint Learning of Salient Object Detection, Depth Estimation and Contour Extraction

IEEE TIP, 2022
Xiaoqi Zhao · Youwei Pang · Lihe Zhang · Huchuan Lu

Motivation - Our High-quality Depth Prediction vs. Previous Low-quality Depth Inputs

Motivation - Depth-free Networks

Pipeline - Multi-task Learning Framework (Depth, Saliency, Contour)

Potential - Predicted Depth Maps on RGB SOD datasets

Potential - Helping Existing Depth-based Methods to Obtain Additional Gains

Datasets

• Google Drive / BaiduYunPan(1a4t)

Trained Models

• MMFT_RES101_duts_njud_nlpr_jointT GitHub Release
• MMFT_RES101_finetune_njud_nlpr GitHub Release
• MMFT_RES250_finetune_njud_nlpr GitHub Release
• MMFT_RES50_duts_njud_nlpr_jointT GitHub Release

Prediction Maps

• Depth_prediction GitHub Release
• Saliency_prediction GitHub Release

Evaluation Tools

• https://github.com/Xiaoqi-Zhao-DLUT/PySegMetric_EvalToolkit
• https://github.com/Xiaoqi-Zhao-DLUT/MMFT/blob/main/Depth_eva.py

Citation

If you think MMFT codebase are useful for your research, please consider referring us:

@article{MMFT,
  title={Joint learning of salient object detection, depth estimation and contour extraction},
  author={Zhao, Xiaoqi and Pang, Youwei and Zhang, Lihe and Lu, Huchuan},
  journal={IEEE Transactions on Image Processing},
  volume={31},
  pages={7350--7362},
  year={2022}
}