mat10599 / streamlit_images_labelling

streamlit_data_labelling

Name

Data labelling app with streamlit.

Description

The goal of this app is to allow the user to label unlabelled images manually.

Installation

First clone the repository on your computer. Then in your terminal either:

1. install the required dependencies running pip install -r requirements.txt (in terminal) or otherwise.
2. run streamlit run labelling_app.py --server.port 8080 (in terminal).

• or

1. run docker build -t name_of_image . (in terminal)
2. run docker run name_of_image (in terminal)
3. type http://0.0.0.0:8502 or http://localhost:8502 in your favourite browser to use the app!

Usage

The app is meant to be used to label images. User should first upload the different images it wants to label. Then, he should state what are the possible labels for the app.Finally, the user can label the different images, review and modify the labels if a mistake was made and download the image names with their labelling.

streamlit-labelling_app-2023-09-16-18-09-55.webm

Roadmap

Future improvements to the app include:

• possibility to connect to a database and to directly "query" unlabelled images from it instead of manually uploading the images.
• "force" the images to be resized in a certain box: bigger images will take a bigger place in the app, shifting elements dynamically at the cost of user experience.

Feedback

Any comment/remark on how I could improve code quality or the app is much appreciated!