File size: 3,811 Bytes
dbe8ac7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
---

license: cc-by-nc-4.0
tags:
- federated-learning
- foundation-models
- medical-imaging
- endoscopy
- self-supervised-learning
- masked-autoencoder
- contrastive-learning
- pytorch

---

# FedFound: Federated Foundation Models for Gastrointestinal Endoscopy

This repository contains pretrained foundation models released as part of the paper:

**FedFound: Federated Foundation Models for Gastrointestinal Endoscopy**

The models were trained using self-supervised learning on gastrointestinal endoscopy images under centralized, local, and federated learning settings. Two pretraining paradigms are provided:

* **Masked Autoencoder (MAE)**
* **Momentum Contrast (MoCo)**

These checkpoints can be used as initialization for downstream gastrointestinal endoscopy tasks such as classification, segmentation, and representation learning.

---

## Available Checkpoints

| Checkpoint             | Clients     | Pretraining |
| ---------------------- | ----------- | ----------- |
| lb_split1.pth          | 1           | MAE         |
| lb_split2.pth          | 1           | MAE         |
| lb_split10.pth         | 10          | MAE         |
| lb_split20.pth         | 20          | MAE         |
| ub_central.pth         | Centralized | MAE         |
| fedavg_split1.pth      | 6           | MAE         |
| fedavg_split2.pth      | 6           | MAE         |
| fedavg_split10.pth     | 10          | MAE         |
| fedavg_split20.pth     | 20          | MAE         |
| fedavgm_split1.pth     | 6           | MAE         |
| fedavgm_split2.pth     | 6           | MAE         |
| fedavgm_split10.pth    | 10          | MAE         |
| fedavgm_split20.pth    | 20          | MAE         |
| fedadam_split1.pth     | 6           | MAE         |
| fedadam_split2.pth     | 6           | MAE         |
| fedadam_split10.pth    | 10          | MAE         |
| fedadam_split20.pth    | 20          | MAE         |
| fedadagrad_split1.pth  | 6           | MAE         |
| fedadagrad_split2.pth  | 6           | MAE         |
| fedadagrad_split10.pth | 10          | MAE         |
| fedadagrad_split20.pth | 20          | MAE         |
| moco_lb_split1.pth     | 1           | MoCo        |
| moco_lb_split2.pth     | 1           | MoCo        |
| moco_ub_central.pth    | Centralized | MoCo        |
| moco_fedavg_split1.pth | 6           | MoCo        |
| moco_fedavg_split2.pth | 6           | MoCo        |

---

## Naming Convention

* **lb**: Lower Bound (single-client training)
* **ub**: Upper Bound (centralized training)
* **fedavg**: FedAvg aggregation
* **fedavgm**: FedAvgM aggregation
* **fedadam**: FedAdam aggregation
* **fedadagrad**: FedAdagrad aggregation
* **moco**: Momentum Contrast (MoCo) pretraining
* Models without the `moco` prefix use Masked Autoencoder (MAE) pretraining

---

## Usage

```python
import torch

checkpoint = torch.load("fedavg_split1.pth", map_location="cpu")

if isinstance(checkpoint, dict) and "model" in checkpoint:
    state_dict = checkpoint["model"]
else:
    state_dict = checkpoint

model.load_state_dict(state_dict, strict=False)
```


## Repository Contents

This repository contains only pretrained model weights.

No patient images, labels, metadata, or clinical information are included.

---

## Citation

If you use these models in your research, please cite:

```bibtex
@article{devkota2025federated,
  title={Federated foundation model for gi endoscopy images},
  author={Devkota, Alina and Amireskandari, Annahita and Palko, Joel and Thakkar, Shyam and Adjeroh, Donald and Jiang, Xiajun and Bhattarai, Binod and Gyawali, Prashnna K},
  journal={arXiv preprint arXiv:2505.24108},
  year={2025}
}
```

---

## Contact

For questions regarding the models, datasets, or training procedures, please open an issue or contact the authors of the paper.