https://github.com/humansignal/smartfew
SmartFew is your swiss knife for semi-supervised structuring of unlabeled data using Few Shot Learning.
https://github.com/humansignal/smartfew
few-shot few-shot-learning few-shot-recognition machine-learning meta-learning one-shot-learning one-shot-segmentation zero-shot zero-shot-learing zero-shot-learning
Last synced: about 1 year ago
JSON representation
SmartFew is your swiss knife for semi-supervised structuring of unlabeled data using Few Shot Learning.
- Host: GitHub
- URL: https://github.com/humansignal/smartfew
- Owner: HumanSignal
- Created: 2019-10-21T12:33:54.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2023-05-22T22:31:01.000Z (about 3 years ago)
- Last Synced: 2023-07-25T05:23:51.460Z (almost 3 years ago)
- Topics: few-shot, few-shot-learning, few-shot-recognition, machine-learning, meta-learning, one-shot-learning, one-shot-segmentation, zero-shot, zero-shot-learing, zero-shot-learning
- Language: Python
- Homepage:
- Size: 1.88 MB
- Stars: 3
- Watchers: 7
- Forks: 3
- Open Issues: 6
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# SmartFew
SmartFew is your swiss knife for semi-supervised structuring of unlabeled data.
#### Install
```bash
pip3 install -e .
```
#### How it works:
1. Prepare the file with image URLs _image_urls.txt_, e.g.
```text
https://myhost.com/image1.jpg
https://myhost.com/image2.jpg
...
```
2. Run server
```bash
cd server && python start.py --input image_urls.txt
```
3. Go to `http://localhost:14321/` in your browser and start selecting relevant images.

Then press **Submit** to continue with a new trial.
The underlying process starts to learn your selection, and you are expecting to see more and more relevant results in your consequent trials.
The algorithm is powered by [Few Shot learning](https://msiam.github.io/Few-Shot-Learning/), that gives an opportunity to learn very fast and quickly adapts to unseen tasks.