# ts-Atlas_APP

**Repository Path**: FreeCode/ts-Atlas_APP

## Basic Information

- **Project Name**: ts-Atlas_APP
- **Description**: 模型炼知框架：构建了炼知平台和重组引擎，通过知识重组图谱实现深度模型可迁移性度量，同时通过重组算法实现新模型生成，重新定义了模型生产方式。
- **Primary Language**: Python
- **License**: Apache-2.0
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 9
- **Created**: 2022-08-03
- **Last Updated**: 2022-08-03

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

<div  align="center">  

<img src="KamalEngine/docs/kae-logo-light.png" width = "30%" height = "30%" />  


# KAmalEngine

</div>

KAmalEngine (KAE) aims at building a lightweight algorithm package for Knowledge Amalgamation and Transferability Estimation. 

Features

  * Knowledge amalgamation and distillation algorithms
  * Easy-to-use Interfaces for multi-tasking training
  * Deep model transferability estimation based on attribution maps
  * Predefined callbacks & metrics for evaluation and visualization

    <img src="KamalEngine/docs/imgs/vis_seg.png" width = "40%" />  
    <img src="KamalEngine/docs/imgs/vis_depth.png" width = "40%" /> 


## Algorithms

### Student Becoming the Master (Task Branching)
Knowledge amalgamation for multiple teachers by feature projection.  
[Student Becoming the Master: Knowledge Amalgamation for Joint Scene Parsing, Depth Estimation, and More](https://arxiv.org/abs/1904.10167) (*CVPR 2019*)  

<img src="KamalEngine/docs/imgs/sbm_results.png"  width = "100%" /> 

### Common Feature Learning
Extract common features from multiple teacher models.  
[Knowledge Amalgamation from Heterogeneous Networks by Common Feature Learning](http://arxiv.org/abs/1906.10546) (*IJCAI 2019*)

Feature Space             |  Common Space
:-------------------------:|:-------------------------:
![cfl-feature-space](KamalEngine/docs/imgs/feature_space_tsne_0.png)  |  ![cfl-feature-space](KamalEngine/docs/imgs/common_space_tsne_0.png)

### Amalgamating Knowledge towards Comprehensive Classification
Layer-wise amalgamation  
[Amalgamating Knowledge towards Comprehensive Classification](https://arxiv.org/abs/1811.02796v1) (*AAAI 2019*)  

<img src="KamalEngine/docs/imgs/layerwise.png"  width = "100%" /> 

### Recombination
Build a new multi-task model by combining & pruning weight matrixs from distinct-task teachers.

<img src="KamalEngine/docs/imgs/recombination.png"  width = "100%" /> 

### Deep model transferability from attribution maps
Estimate model transferability using attribution map.

<img src="KamalEngine/docs/imgs/attrmap.png"  width = "100%" /> 

### DEPARA: Deep Attribution Graph for Deep Knowledge Transferability
Constructing attribution graph for model transferability estimation.

<img src="KamalEngine/docs/imgs/attrgraph.png"  width = "100%" /> 

<figure>
<img src="KamalEngine/docs/imgs/transgraph.png"  width = "100%" /> 
<figcaption align="center">Transferability graph on classification models</figcaption>
</figure>

## Team

Developed by [Zhejiang Lab](http://www.zhejianglab.com/) and [VIPA Lab](http://vipazoo.cn) from Zhejiang University.

## Citation
```
@inproceedings{shen2019amalgamating,
  author={Shen, Chengchao and Wang, Xinchao and Song, Jie and Sun, Li and Song, Mingli},
  title={Amalgamating Knowledge towards Comprehensive Classification},
  booktitle={AAAI Conference on Artificial Intelligence (AAAI)},
  pages={3068--3075},
  year={2019}
}
```

```
@inproceedings{ye2019student,
  title={Student Becoming the Master: Knowledge Amalgamation for Joint Scene Parsing, Depth Estimation, and More},
  author={Ye, Jingwen and Ji, Yixin and Wang, Xinchao and Ou, Kairi and Tao, Dapeng and Song, Mingli},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  pages={2829--2838},
  year={2019}
}
```

```
@inproceedings{luo2019knowledge,
  title={Knowledge Amalgamation from Heterogeneous Networks by Common Feature Learning},
  author={Luo, Sihui and Wang, Xinchao and Fang, Gongfan and Hu, Yao and Tao, Dapeng and Song, Mingli},
  booktitle={Proceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI)},
  year={2019},
}
```

```
@inproceedings{shen2019customizing,
  author={Shen, Chengchao and Xue, Mengqi and Wang, Xinchao and Song, Jie and Sun, Li and Song, Mingli},
  title={Customizing student networks from heterogeneous teachers via adaptive knowledge amalgamation},
  booktitle={The IEEE International Conference on Computer Vision (ICCV)},
  year={2019}
}
```

```
@inproceedings{Ye_Amalgamating_2019,
  year={2019},
  author={Ye, Jingwen and Wang, Xinchao and Ji, Yixin and Ou, Kairi and Song, Mingli},
  title={Amalgamating Filtered Knowledge: Learning Task-customized Student from Multi-task Teachers}
  booktitle={Proceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI)},
  year={2019},
}
```

```
@inproceedings{song2020depara,
  title={DEPARA: Deep Attribution Graph for Deep Knowledge Transferability},
  author={Song, Jie and Chen, Yixin and Ye, Jingwen and Wang, Xinchao and Shen, Chengchao and Mao, Feng and Song, Mingli},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={3922--3930},
  year={2020}
}
```

```
@inproceedings{song2019deep,
  title={Deep model transferability from attribution maps},
  author={Song, Jie and Chen, Yixin and Wang, Xinchao and Shen, Chengchao and Song, Mingli},
  booktitle={Advances in Neural Information Processing Systems},
  pages={6182--6192},
  year={2019}
}
```