SMART Swipe - A fast labelling tool for AI image datasets

Author : Pi Ko (pk2269@nyu.edu)

SMART Swipe is a fast labelling tool for AI image datasets written in Flutter. It runs locally without any dependency, with an intuitive interface.

Running the Software

First download this entire repository to your computer.

Then just run the built linux binary in

./build/linux/x64/release/bundle/smart_swipe_local 

Supported Image Formats : 'png', 'jpg', 'jpeg', 'gif', 'bmp', 'webp'

⚠️ IMPORTANT : If the original image file size is too large, the software may not be able to load the image. In that case, please resize the image to a smaller size before loading.

Key Bindings: To make the labelling process quicker, the user can use I and P keys to navigate between the images, and Y and N to label the image as Yes or No.

After labelling, the labels are saved as text file with the same name as image file in the same directory as the image folder.

Modifying the Software

Development Setup

To edit and compile the binary, follow the flutter installation tutorial for Linux at https://docs.flutter.dev/get-started/install/linux

Then get the Flutter extension on Visual Studio Code, and open the project folder.

Then, install all the dependencies by running

flutter pub get

Then, you can run the app on your device by running

flutter run

Building Binary

Add Desktop support to the Flutter App by Running

flutter create --platforms=windows,macos,linux .

Then clear previous build artifacts by running

flutter clean

To build the binary, run

flutter build linux

After building, the built linux binary is in

./build/linux/x64/release/bundle/smart_swipe_local 

License

This tool is released under the MIT License. See LICENSE for more details.

About

This tool has been developed as a collaboration between Pi and S.M.A.R.T. Construction Research Group at New York University Abu Dhabi.

Please contact pk2269@nyu.edu for any issues regarding this software.