---
language: 
  - en
tags:
- background-removal
license: apache-2.0
library: pytorch
model-index:
- name: IS-Net_DIS-general-use
  results:
  - task:
      name: Image Segmentation
      type: image-segmentation
    metrics:
    - name: Human Correction Efforts
      type: human_correction_efforts
      value: 1016
      
---

# IS-Net_DIS-general-use

* Model Authors: Xuebin Qin, Hang Dai, Xiaobin Hu, Deng-Ping Fan*, Ling Shao, Luc Van Gool
* Paper: Highly Accurate Dichotomous Image Segmentation (ECCV 2022 - https://arxiv.org/pdf/2203.03041.pdf
* Code Repo: https://github.com/xuebinqin/DIS
* Project Homepage: https://xuebinqin.github.io/dis/index.html

Note that this is an _optimized_ version of the IS-NET model.

From the paper abstract:

> [...]  we introduce a simple intermediate supervision baseline (IS- Net) using both feature-level and mask-level guidance for DIS model training. Without tricks, IS-Net outperforms var- ious cutting-edge baselines on the proposed DIS5K, mak- ing it a general self-learned supervision network that can help facilitate future research in DIS.

![](https://raw.githubusercontent.com/xuebinqin/DIS/main/figures/is-net.png)

# Citation

```
@InProceedings{qin2022,
      author={Xuebin Qin and Hang Dai and Xiaobin Hu and Deng-Ping Fan and Ling Shao and Luc Van Gool},
      title={Highly Accurate Dichotomous Image Segmentation},
      booktitle={ECCV},
      year={2022}
}
```