https://github.com/dnth/self-supervised-timm
Adapt TIMM backbone to your custom dataset with self-supervised learning in fastai
https://github.com/dnth/self-supervised-timm
Last synced: over 1 year ago
JSON representation
Adapt TIMM backbone to your custom dataset with self-supervised learning in fastai
- Host: GitHub
- URL: https://github.com/dnth/self-supervised-timm
- Owner: dnth
- License: apache-2.0
- Created: 2024-07-23T12:54:01.000Z (almost 2 years ago)
- Default Branch: main
- Last Pushed: 2024-12-31T15:11:34.000Z (over 1 year ago)
- Last Synced: 2025-03-25T18:12:25.168Z (over 1 year ago)
- Language: Jupyter Notebook
- Size: 18.1 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# self-supervised-timm
Adapt TIMM backbone to your custom dataset with self-supervised learning and convert model into ONNX format for deployment.
## Why?
TIMM has a wide range of pre-trained models that can be used for transfer learning. However, the pre-trained models are trained on ImageNet dataset which may not be suitable for your custom dataset.
This repository provides a way to adapt TIMM backbone to your custom dataset with self-supervised learning and convert the model into ONNX format for deployment.
## How?
Using self-supervised learning (SSL) we adapt the TIMM backbone to the custom dataset without any labels. Once the SSL training is done, we can optionally fine-tune the model with labels. Finally, we convert the model into ONNX format for deployment.
## Dataset
MVTec AD Zipper Images - https://www.kaggle.com/datasets/atthaariq/resized-zipper
## Setting Up
Make sure you have PyTorch installed. Next, install the required packages.
```bash
pip install fastai timm onnx self-supervised
```