FRoundation

KO | EN

gitlite — search

#python #ai #golang #android #deep-learning #linux #nodejs #java #awesome #cli #kubernetes #awesome-list #pytorch #ios

(★ 14)

FRoundation: Are Foundation Models Ready for Face Recognition?

FRoundation 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# FRoundation: Are Foundation Models Ready for Face Recognition?

Tahar Chettaoui,
Naser Damer,
Fadi Boutros

[[`Paper`](https://arxiv.org/abs/2410.23831)] [[`Blog`](https://taharchettaoui.github.io/FRoundation_web/)]

PyTorch implementation and pretrained models for **FRoundation**.

### Pretrained models

| Method | Backbone | Train data | Link |
| -------- | ------- | ------- | ------- |
| Baseline | Small Large | CASIA-WebFace MS1MV2 WebFace4M CASIA-WebFace MS1MV2 WebFace4M | [Download](https://owncloud.fraunhofer.de/index.php/s/RqDWPal8qvXKuH7) [Download](https://owncloud.fraunhofer.de/index.php/s/7HubejKcXwhxk8D) [Download](https://owncloud.fraunhofer.de/index.php/s/ZXnzAODROxWncil) [Download](https://owncloud.fraunhofer.de/index.php/s/bugYxj2g6lmbtbZ) [Download](https://owncloud.fraunhofer.de/index.php/s/AeaxyTB1j6U2Nb5) [Download](https://owncloud.fraunhofer.de/index.php/s/XGFtvGLFM39aPtV)|
| CLIP | Base Large | CASIA-WebFace MS1MV2 WebFace4M CASIA-WebFace MS1MV2 WebFace4M|[Download](https://owncloud.fraunhofer.de/index.php/s/oeyZTsXKYKFID5M) [Download](https://owncloud.fraunhofer.de/index.php/s/OyD0N0KYvyToBBr) [Download](https://owncloud.fraunhofer.de/index.php/s/p2ZbWOsp1nVLv0f) [Download](https://owncloud.fraunhofer.de/index.php/s/tkz0g8og3pKpISd) [Download](https://owncloud.fraunhofer.de/index.php/s/fYKGSSA1HviGIBo) [Download](https://owncloud.fraunhofer.de/index.php/s/w9w0Sa4vnDYd79y)|

### Usage
First, install all the dependencies listed in the requirements.txt file. This code has been tested with Python 3.9. Below is an example of how to load the model and pretrained weights, specifically for CLIP ViT-B/16.

```
import sys
import os
import torch
import torch.distributed as dist

from datetime import timedelta
from torch.nn.parallel import DistributedDataParallel

from backbone import get_model
from finetuning import apply_lora_model
from data.transform import transform_image
from utils.evaluation import CallBackVerification

# Load pretrained model
model = get_model(rank, model_name="clip", backbone_size='ViT-B/16')

# Attach LoRA layers
apply_lora_model(
    rank, 
    model, 
    training_type="image_encoder_only",
    model_name="clip",
    backbone_size='ViT-B/16', 
    lora_target_modules=['q', 'v'],
    lora_r=16, 
    lora_a=32, 
    lora_dropout=0.25, 
    device=rank, 
    position="all"
)

# Load pretrained model
model_path = "path_to_pretrained_model" 
model.backbone.load_state_dict(torch.load(model_path))
model.backbone.visual # To access the image encoder of CLIP
```

### Training
After updating the config.config.py file with the required paths and desired parameters, you can initiate training using:

```
./train.sh
```

In addition to logs that document the loss and hyperparameters during training, a **TensorBoard** with various plots is also available.

### Coming soon ...
- [ ] Pretrained models link
- [ ] Evaluation code (IJB-C / IJB-B)

### Citation

```
@article{DBLP:journals/ivc/ChettaouiDB25,
  author       = {Tahar Chettaoui and
                  Naser Damer and
                  Fadi Boutros},
  title        = {FRoundation: Are foundation models ready for face recognition?},
  journal      = {Image Vis. Comput.},
  volume       = {156},
  pages        = {105453},
  year         = {2025}
}
```

### License

```
This project is licensed under the terms of the Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) license. 
Copyright (c) 2021 Fraunhofer Institute for Computer Graphics Research IGD Darmstadt
```

FRoundation

FRoundation 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

FRoundation

FRoundation 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README